ACS Publications. Most Trusted. Most Cited. Most Read
Quick View

OR SEARCH CITATIONS

My Activity
Publications
CONTENT TYPES

Figure 1Loading Img
Download Hi-Res ImageDownload to MS-PowerPointCite This:J. Proteome Res. 2021, 20, 7, 3489-3496
RETURN TO ISSUEPREVArticleNEXT

DIMA: Data-Driven Selection of an Imputation Algorithm

  • Janine Egert*
    Janine Egert
    Institute of Medical Biometry and Statistics (IMBI), Institute of Medicine and Medical Center Freiburg, 79104 Freiburg im Breisgau, Germany
    Centre for Integrative Biological Signalling Studies (CIBSS), Albert-Ludwigs-Universität Freiburg, 79104 Freiburg, Germany
    *Email: [email protected]
    More by Janine Egert
    Orcidhttps://orcid.org/0000-0002-2032-7081
  • , 
  • Eva Brombacher
    Eva Brombacher
    Institute of Medical Biometry and Statistics (IMBI), Institute of Medicine and Medical Center Freiburg, 79104 Freiburg im Breisgau, Germany
    Centre for Integrative Biological Signalling Studies (CIBSS), Albert-Ludwigs-Universität Freiburg, 79104 Freiburg, Germany
    Spemann Graduate School of Biology and Medicine (SGBM), Albert-Ludwigs-Universität Freiburg, 79104 Freiburg, Germany
    Faculty of Biology, Albert-Ludwigs-Universität Freiburg, 79104 Freiburg im Breisgau, Germany
    More by Eva Brombacher
  • , 
  • Bettina Warscheid
    Bettina Warscheid
    Biochemistry and Functional Proteomics, Institute of Biology II, Faculty of Biology, Albert-Ludwigs-Universität Freiburg, 79104 Freiburg im Breisgau, Germany
    Signalling Research Centres BIOSS and CIBSS, Albert-Ludwigs-Universität Freiburg, 79104 Freiburg im Breisgau, Germany
    More by Bettina Warscheid
    Orcidhttps://orcid.org/0000-0001-5096-1975
  • , and 
  • Clemens Kreutz*
    Clemens Kreutz
    Institute of Medical Biometry and Statistics (IMBI), Institute of Medicine and Medical Center Freiburg, 79104 Freiburg im Breisgau, Germany
    Signalling Research Centres BIOSS and CIBSS, Albert-Ludwigs-Universität Freiburg, 79104 Freiburg im Breisgau, Germany
    Center for Data Analysis and Modeling (FDM), Albert-Ludwigs-Universität Freiburg, 79104 Freiburg im Breisgau, Germany
    *Email: [email protected]
    More by Clemens Kreutz
Cite this: J. Proteome Res. 2021, 20, 7, 3489–3496
Publication Date (Web):June 1, 2021
https://doi.org/10.1021/acs.jproteome.1c00119

Copyright © 2022 The Authors. Published by American Chemical Society. This publication is licensed under

CC-BY-NC-ND 4.0.
  • Open Access

Article Views

1356

Altmetric

-

Citations

3
LEARN ABOUT THESE METRICS
PDF (3 MB)
Get e-Alerts
Supporting Info (1)»
SUBJECTS:

Abstract

High Resolution Image
Download MS PowerPoint Slide

Imputation is a prominent strategy when dealing with missing values (MVs) in proteomics data analysis pipelines. However, it is difficult to assess the performance of different imputation methods and varies strongly depending on data characteristics. To overcome this issue, we present the concept of a data-driven selection of an imputation algorithm (DIMA). The performance and broad applicability of DIMA are demonstrated on 142 quantitative proteomics data sets from the PRoteomics IDEntifications (PRIDE) database and on simulated data consisting of 5–50% MVs with different proportions of missing not at random and missing completely at random values. DIMA reliably suggests a high-performing imputation algorithm, which is always among the three best algorithms and results in a root mean square error difference (ΔRMSE) ≤ 10% in 80% of the cases. DIMA implementation is available in MATLAB at github.com/kreutz-lab/OmicsData and in R at github.com/kreutz-lab/DIMAR.

This publication is licensed under

CC-BY-NC-ND 4.0.
  • cc licence
  • by licence
  • nc licence
  • nd licence
KEYWORDS:

1. Introduction

ARTICLE SECTIONS
Jump To
Mass spectrometry (MS) has become a popular strategy for identifying and quantifying thousands of proteins in proteomics research. Despite the ongoing technical improvements in modern MS technologies, the handling of missing values (MVs) remains a persistent issue (1−6) as high rates of MVs─with up to 30–50% MVs─- result in missing identification and missing quantification.
The occurrence of missing data results from different biological and/or technical reasons: a peptide is either not detected or falsely identified (missing completely at random (MCAR)), below the detection limit (missing not at random (MNAR)) or simply not present in the sample. (6−8)
A further consequence of MVs is that they bias estimations and may negatively impact statistical power and downstream analyses such as statistical tests and clustering analyses. (9−11) Many of these even require complete data.
A popular strategy to deal with incomplete data is to apply imputation. (5,6,8,9) However, a poor-performing imputation method may bias subsequent analysis steps making missing value analysis an ongoing cause for debate within the bioinformatics community. To date, there is no “one fits all” imputation strategy available. Many imputation methods and method comparisons specifically tailored to a certain data acquisition technique have been published in the last few years. (12−19) Furthermore, serial dilution data were used to find a generally superior imputation strategy; (5,20,21) however, no such strategy was identified. (5,12,22)
Recently, the tool, NAguideR, (23) was developed to guide the decision for an optimal data-dependent imputation algorithm. It compares various imputation strategies on the complete portion of the data based on which an optimal imputation method for the incomplete portion is suggested. However, in general, only 10–30% of the quantified proteins have no MV, so the complete portion may not reflect the whole data set. In addition, the random assignment of MVs may not be a realistic representation of the occurrence of MVs.
With data-driven selection of an imputation algorithm (DIMA) we extend this approach to a more realistic setting. DIMA provides a valid recommendation of a high-performing imputation algorithm specifically tailored to the user-supplied data set and its MV distribution. Within DIMA, a reference data set is generated that captures the intensity and MV distribution from the original data and on which various imputation strategies can be evaluated. The reference data comprises proteins with the least amount of MVs scaled to the protein means and standard deviations of the whole input data set. To reproduce the pattern of MVs, a logistic regression model is trained on the original data taking into account the protein and sample as well as the mean protein intensity. The logistic regression coefficients are then applied to the reference data to imitate the MV distribution.

2. Methods

ARTICLE SECTIONS
Jump To

2.1. Illustration Data

To illustrate the concept of DIMA, it is applied to exemplary liquid chromatography (LC)–mass spectrometry (MS)/MS data set. Here, the secretion of antitumoral factors by rat alveolar wild-type macrophages is triggered by four different concentrations of proprotein convertase (PC) inhibitors ({0, 50, 100, 150} μM) leading to a decreased viability of C6 glioma cells (PXD014679/proteinGroups-WTsecretome). (24) This is measured in three technical replicates. The data comprises 884 identified proteins with 39.3% MVs and is depicted in Figure 1.

Figure 1

Figure 1. DIMA analysis pipeline illustrated on an LC-MS/MS data set. (24) The data is sorted from top to bottom according to the frequency of MVs and the mean intensity of the proteins. Likewise, the reference data R is sorted according to the mean protein intensities after considering pattern PR of step 3. (1) The pattern PO of MVs is learned by logistic regression using the protein and sample as factorial predictors and the mean protein intensity as a continuous predictor. (2) A reference data R with few MVs is defined. (3) Various patterns PR of MVs are generated by the logistic regression model, and the respective coefficients of step 1 are incorporated into the reference data R. (4) Boxplots of the absolute imputation errors for multiple imputation algorithms. The circle indicates the median imputation deviation. The algorithms are ranked by their overall root mean square error (RMSE, red diamond). The algorithms can be divided into well-performing algorithms with an RMSE < 0.5 (green), medium performance with 0.5 < RMSE < 3 (yellow), and bad performance with RMSE > 3 (red). (5) The best-performing imputation algorithm on R (in this example impSeqRob) is recommended for the original data O and imputation of O is conducted.

High Resolution Image
Download MS PowerPoint Slide

2.2. PRIDE Data

To show its broad applicability and performance, DIMA is applied to 142 publicly available data sets from the PRoteomics IDEntifications (PRIDE) database. Data sets that contain the search pattern “MaxQuant” (25) or “proteinGroups”, the file extensions .txt, .xlsx, .csv, .tar, .gz, .zip or .rar and uploaded to the PRIDE database between May and July 2020 are acquired from the ftp-server “ftp.PRIDE.ebi.ac.uk”. The resulting 142 PRIDE data sets comprise [200–13 430] proteins, [2–112] samples and [2.6–94.3]% MVs, with an average amount of (2990 ± 200) proteins, (17.8 ± 1.3) samples, and (40.5 ± 2.1)% MVs. Thirty-five percent of the data sets contain more than 50% MVs. The characteristics for the individual PRIDE data sets are shown in the Supporting Information Table S2.

2.3. Simulation Study

In addition to experimental data sets, a simulation study is performed to analyze the influence of MCAR and MNAR ratios as well as the frequency of MVs on the imputation performance. The simulation process is adapted from O’Brien et al. (26) The protein intensities
(1)
of condition k are calculated with the baseline average θij ∼ N(18.5, IG(1,1)) of peptide j matched to protein O and with a noise term ϵijk ∼ N(0, IG(2, 1)). To simulate differential expression, a fold change FCik ∼ N(0, IG(1.5,1)) is added to the protein intensity for condition k = 1. The variances are drawn from the inverse γ distribution IG, which represents the marginal posterior distribution for unknown variances of a normal distribution.
To evaluate the impact of the MV distribution, various combinations of MVs and MCAR/MNAR ratios are incorporated into the simulated data Y. The pattern Pik ∈ {0, 1} of MVs is defined as Pik = 1 if the observation is available and Pik = 0 if the observation is missing. The MCAR values are set randomly. The indicator variable
(2)
of MNAR values is drawn from the intensity-dependent probit model Φ(a + bYik) with the cumulative distribution function Φ() of a normally distributed random variable N(0, 1), (26) the rate b of MNAR values; and the b-quantile a = Qb(Y) of the protein intensities Y. In our notation, MVs are incorporated in the simulated data S if one of the patterns PikMNAR ∥ PikMCAR = 0 suggests an MV, otherwise, S consists of the simulated intensities Y. For each MV and MCAR/MNAR ratio, 500 data sets with 500 proteins and 20 samples that consist of two conditions k = {0, 1} are simulated.

2.4. DIMA

DIMA assesses and suggests imputation algorithms for a user-defined data set. The method consists of five main steps, which are depicted in Figure 1.
(1) The pattern PO of MVs in the original input data O is learned by logistic regression (Section 2.4.1).
(2) Reference data R with the subset of proteins containing the least amount of MVs are generated from the original data O to evaluate imputation performance (Section 2.4.2).
(3) To generate a pattern of missing data with a similar distribution as in the original data, the logistic regression model of step 1 is applied to the reference data R. Bernoulli trials are performed to simulate nP patterns PR of MVs (Section 2.4.2).
(4) Multiple imputation algorithms (Section 2.4.3) are applied to R with patterns PR of MVs and ranked by their root mean square error (RMSE). The best-performing algorithm is given by the lowest average rank over all pattern simulations (Section 2.4.4).
(5) The best-performing imputation algorithm of step 4 is recommended as an imputation algorithm for the original data O and imputation of O is performed.
In general, as the protein abundances are log-normally distributed, the authors recommend applying DIMA on the logarithmic scale. In this paper, intensities are transformed into a logarithmic scale with base 2.

2.4.1. Learn Pattern of Missing Values

The probability pMV = Prob(P = 0) of a specific data point missing (P = 0) is described by a logistic regression model as follows
(3)
The columns of the design matrix X correspond to predictor variables describing certain data properties: the mean protein intensity that is central for MNAR data and the protein and sample identities as factorial predictors. The sample predictor captures different experimental conditions and replicates. In addition, DIMA offers optional predictors such as protein ratios, molecular weight, or quality scores if included in the input file.
The predictors are standardized (27,28) and a weak regularization of the regression coefficients β for the factorial predictors has been performed to decrease the variance of the estimated parameters and to prevent non-identifiability. (29) For large data sets (>1000 proteins), multiple random subsamples of the proteins are analyzed to decrease the computational cost. Here, each protein is drawn once and the regression coefficients β are set to the sampling means.

2.4.2. Generation of the Reference Data

Because the imputation algorithms are evaluated and ranked on the reference data R, the reference data plays a key role within DIMA. The aim is to simulate abundances, variability, and MV distribution as realistically as possible compared to the original data O. This is done by assigning the proteins with the least number of MVs─at least 20% of the proteins─to the reference data multiple times to get the same number of proteins as in the original data. The reference data is then scaled to the mean and standard deviation of the original proteins. To simulate the occurrence patterns PR of MVs, the logistic model from eq 3 with the regression coefficients β of step 1 is applied to the reference data R, and Bernoulli trials are performed. Depending on the size of the input data, 5–20 patterns are generated.

2.4.3. Imputation Algorithms

By applying DIMA, 30 imputation algorithms from 12 R packages are evaluated. They are based on a variety of imputation strategies such as single-value approaches (mean, minimum), local similarity approaches (KNN, random forest, regression models), and global structure approaches (sequential, PCA, SVD). The algorithms are depicted in the Supporting Information, Table S1, with their package name, reference, and a short explanation.

2.4.4. Ranking of Imputation Algorithms

To compare the performance of the different imputation algorithms, the root mean square error
(4)
between the protein intensities Ri of the reference data and the imputed protein intensities Ii is calculated. Here, n denotes the total number of imputed entries.
The imputation algorithms are ranked by their RMSE for each generated pattern. The approach with the lowest mean rank over all pattern simulations is recommended as the best-performing algorithm for the original data O. If for any reason an imputation algorithm fails, the highest rank is assigned for this respective imputation. Thus, the algorithm may still be recommended by DIMA, even though this is unlikely. An imputed data set that still contains MVs is by definition treated as an imputation failure.
Most commonly, the accuracy of the imputed point estimates to the underlying truth is desired for downstream analyses including estimation of the protein fold changes or cluster analysis. For statistical tests, however, also the variability of a data set is crucial and it might not be reproduced by the imputed point estimates. In such a setting, alternative criteria have to be applied to rank the imputation algorithms. For statistical interpretation based on the t-test, the RMSEt ≔ RMSE(tR, tI) serves as rank criterion, where t is the t-test statistics calculated from the observed data R and the imputed data O. To define the null hypothesis H0, the group assignments of the samples have to be specified by the user.
To verify that the selected imputation approach preserves the variability of the data, the differences in variances of the observed data R and the imputed data I are examined with a two-sample F-test. Here, the p-value
(5)
under H0 of equal variances σR2 = σI2 is the criterion of choice.

3. Implementation

ARTICLE SECTIONS
Jump To
DIMA is implemented in MATLAB (30) and R (31). The MATLAB implementation is part of the OmicsData repository at github.com/kreutz-lab/OmicsData and the R implementation can be found at github.com/kreutz-lab/DIMAR. All results of the following sections were performed in MATLAB on an Intel Xeon E5-2640v3 CPU with 16 cores on the BwForCluster MLS&WISO Development.

3.1. MATLAB Implementation

A proteomics data object is created by
O = OmicsData(file);
Accepted input formats are .xls, .xlsx, .csv, .tsv, .txt, and .mat files as well as numeric data matrices. A pre-processing step may be applied by
O = OmicsFilter(O,nacut,logflag,scaleflag);
This removes features with a higher MV proportion than the nacut threshold. The flags logflag and scaleflag can be logical or string inputs. They define if a log 2 or log 10 transformation and a median or mean normalization should be performed. DIMA is executed via
O = DIMA(O,[algorithms],[bio]);
By default, the imputation algorithms of Table S1 are applied. Alternatively, they can be specified by the user. A fast version, which only runs the nine most frequently recommended algorithms based on the 142 PRIDE data sets, is also implemented and available by setting the argument algorithms to “fast”. The optional third input argument is a flag if biological information such as protein ratios, molecular weight, or scores should be taken into account. After applying DIMA the suggested algorithm and the respective imputation are stored in the proteomics data object and can be accessed via
algorithm = get(O,’DIMA’);
data = get(O,’data’);

3.2. R Implementation

The DIMA implementation in R is located in the dimar package:
devtools::install_github(”kreutz-lab/DIMAR”)
library(dimar)
The algorithm is applied to the input data mtx in a single function call:
Imp <- DIMAR::dimar(mtx, pattern=NULL, methods=’fast’).
mtx should be given in a matrix format. The MaxQuant output file names can be passed to the function as string, such as the proteinGroups text file. DIMA is applied to those columns of the data whose column name matches the argument pattern. The imputation algorithms can be defined in methods. By default, the nine most frequently selected algorithms on the 142 Pride data sets are applied.

4. Results

ARTICLE SECTIONS
Jump To
DIMA is first demonstrated on one of the PRIDE data sets. Then, to show its broad range of applicability and to evaluate its performance, DIMA is applied to 142 experimental data sets from the PRIDE database (Section 2.2) and to the simulated MNAR and MCAR data sets (Section 2.3).

4.1. Illustration Data

Figure 1 illustrates the DIMA workflow on the LC-MS/MS illustration data set. The protein abundances and MV distribution in the reference data R and the input data O are of great similarity (see the Supporting Information, Figures S1 and S2, for more details). To evaluate various imputation algorithms, the absolute deviations of the imputed data from the original values are calculated and the algorithms are sorted according to their RMSE (Figure 1.4). The performance measures RMSE, RMSEt of the t-statistics, and the p-value (pF) of the F-statistics with their respective ranking are depicted in Table 1 for the eight best-performing and the two least-performing imputation algorithms.
Table 1. To compare the imputed data values to the reference data, the RMSE, RMSE of the t-statistics and p-value of the F-statistics as well as their respective rankings are calculateda
a

The eight best-performing and the two least-performing imputation algorithms on the illustration data set are shown. In green, the best-performing algorithms for the respective criterion are highlighted with decreasing transparency. The algorithm selection by DIMA depending on the ranking criterion is highlighted in bold.

The best-performing algorithm for this exemplary data set is the sequential algorithm impSeqRob from the R package rrocovNA (32) with an average rank of 1.5 over all pattern simulations, followed by the algorithm impSeq from the same R package. The four best-performing algorithms impSeqRob and impSeq as well as the PCA-based algorithms imputePCA and MIPCA from the R package missMDA (33) perform relatively well for all three ranking criteria, which can further be generalized for the remaining investigated PRIDE data sets (Section 2.2). The other imputation algorithms result in higher RMSEs.
The computation time of the individual imputation algorithms ranges from 3 to 120 s. DIMA applied to the illustration data set takes 18 min in total. All imputation methods show a positive correlation between the original and the imputed data values with a Pearson correlation coefficient between 0.59 and 0.99. Most algorithms overestimate, but the R package imputeLCMD which algorithms are left-censored underestimate intensities (Supporting Information Figure S5).
To determine if the protein intensities with and without the treatment of cells with 150 μM proprotein convertase (PC) inhibitor differ significantly from each other, the t-test is applied. The RMSE of the t-statistics over all identified proteins is the smallest for the imputation algorithm imputePCA. Hence, imputePCA is suggested by DIMA for statistical testing. The null hypothesis of equal variances of the original and the imputed data is rejected (pF < 0.01) for 18 and not rejected for 5 out of 23 imputation algorithms. In our example, those 5 algorithms also perform best with respect to the RMSE and RMSEt.
Proteins with complete missingness are also reflected in the patterns PR of MVs in the reference data and thus play a role in the evaluation of the imputation algorithms. However, the R packages pcaMethods and impute are not able to deal with complete missingness. If such proteins are present in the input data, the user can consider removing them from the data beforehand.
The illustration data set comprises the proteins of the WT secretome experiment. (24) The reference data and simulated patterns of MVs for the peptide data of the same experiment are shown in the Supporting Information Figure S4. In principle, DIMA can be applied to any data in a matrix format.

4.2. PRIDE Data

To demonstrate its general applicability and to evaluate the various imputation algorithms on multiple data sets, DIMA is applied to 142 data sets from the PRIDE database (see Section 2.2). For 47% of the data sets, DIMA proposes the robust sequential algorithm impSeqRob and for 25% the sequential algorithm impSeq (Figure 2A). Both approaches are included in the rrcovNA package and are based on sequential imputation of each MV by minimizing the determinant of the covariance matrix. Furthermore, frequently selected imputation algorithms are the random forest algorithm missForest (13%) and the PCA algorithms imputePCA (10%), ppca (1.5%), and bpca (1.5%).

Figure 2

Figure 2. DIMA is applied and evaluated on 142 PRIDE data sets. (A) Nine algorithms compete for being recommended as the best-performing algorithm. The R package rrcovNA with its algorithms impSeqRob (47%) and impSeq (25%) is selected most frequently, followed by missForest in 13% and imputePCA in 10%. For 5% of the Pride data sets, another algorithm is suggested. (B) The rank of the imputation algorithms obtained in the 142 PRIDE data sets is shown as a box plot. The seven algorithms with the lowest median rank are also the seven most frequently selected algorithms by DIMA (A). The algorithms with a median rank lower than 5% are highlighted in green, and algorithms with a median rank greater than 20 are highlighted in red.

High Resolution Image
Download MS PowerPoint Slide
The algorithms most frequently selected by DIMA also show generally good performance on the 142 PRIDE data sets. The eight proposed algorithms have the lowest median rank, except for the algorithm softImpute, which has a median rank of 28 and is the best-performing algorithm for just one data set (Figure 2B). The algorithms impSeqRob and impSeq from the R package rrcovNA show the best performance with a median rank of 2, followed by the R package missMDA (imputePCA and MIPCA) and missForest. On average, DIMA takes 6.44 ± 0.06 min CPU time per data set.
To show the broad applicability of DIMA, in addition to the PRIDE data sets processed using MaxQuant, DIMA is applied to the quantitative output of other proteomics and metabolomics software. Khoonsari et al. (34) identified the peptides of the cerebrospinal fluid proteome and analyzed the same raw data with five different proteomics software: DecyderMS (GE healthcare), MaxQuant, (25) OpenMS, (35) PEAKS (Bioinformatics Solutions Inc.), and Sieve (Thermo). DIMA is applied to all five data sets normalized to the spiked-in chicken ovalbumin. In this example, irrespective of the applied proteomics software, the same imputation algorithm, the sequential algorithm impSeq, is selected as the best-performing imputation algorithm. Thus, in this case, the kind of proteomics software does not seem to be decisive for the selection of the imputation algorithm. For PXD002099 (36) and PXD002885 (37) processed with Progenesis QI (Nonlinear Dynamics, Waters, U.K.) DIMA selected the robust sequential algorithm impSeqRob and the Bayesian principal component analysis (bpca) algorithm, respectively. For the metabolomics MTBLS738 (38) data set DIMA selected the random forest algorithm missForest as the best-performing imputation method.

4.3. Simulation Study

To investigate whether DIMA is capable of reliably identifying a high-performing imputation approach, a data simulation study is conducted where the ground truth for the intensities of MVs is known. The data is simulated as described in Section 2.3 with MV ∈ {5, 10, 15, 20, 25, 30, 35, 40, 45, 50} % and MNAR ∈ {0, 10, 20, 30, 40, 50, 60, 70, 80, 90, 100}%.
First, the simulated data sets S with incorporated MVs are imputed and compared to the simulated ground truth Y. This assessment and the respective ranking of the methods are hereafter referred to as direct imputation assessment. In such a setting, the aim of DIMA is to predict a high-performing method of this direct imputation assessment only based on S. Therefore, the direct imputation assessment serves as the assumed true algorithm ranking against which the recommendation of DIMA is compared.
The RMSE difference of the algorithm recommended by DIMA compared to direct imputation assessment is averaged over 500 data simulations and is displayed in Figure 3. This is shown for different combinations of MV percentages and MNAR/MCAR ratios. The mean RMSE difference is below 5% in 60% of the MV and MNAR/MCAR combinations.

Figure 3

Figure 3. Performance of DIMA is evaluated on simulated data S with the incorporation of various proportions of MV and MNAR/MCAR ratios. The RMSE (color-coded) and rank (first entry) obtained by the best-performing imputation algorithm recommended by DIMA (second entry) compared to direct imputation assessment (third entry) over 500 data simulations are calculated. The algorithm recommended by DIMA is within the top three out of 27 approaches in all cases. For MV < 20% (A), the additive regression aregImpute with type regression (reg) outperforms, between 20 and 30% MVs; (B) several algorithms compete against each other and for MV > 30%; (C) the random forest algorithm missForest performs best.

High Resolution Image
Download MS PowerPoint Slide
In each image section of Figure 3, the rank (first entry) of the algorithm recommended by DIMA (second entry) compared to the best-performing algorithm with direct imputation assessment (third entry) is averaged over all simulated data sets S. The average rank is between 1.2 and 2.2 indicating that DIMA recommends one of the best three out of 27 algorithms independent of the number of MVs or the MNAR/MCAR ratio. The most frequent algorithm recommendations by DIMA and by direct imputation assessment are as follows: For MV < 20% (Figure 3A) the additive regression algorithm aregImpute with type regression from the R package Hmisc. Between an MV percentage of 15 and 35% (Figure 3B) the algorithms aregImpute, impSeq, and missForest compete as the best-performing imputation algorithm depending on the simulated data set. For MV > 30% (Figure 3C) the random forest algorithm missForest is most frequently recommended.

5. Discussion

ARTICLE SECTIONS
Jump To
Despite enormous progress, modern MS-based proteomics may still suffer from the presence of MVs (1−3) and no general guidelines on how to deal with MVs exist. (5,12,22) Poor-performing MV imputation methods as well as performing no imputation at all may lead to an estimation bias and negatively affect the peptide/protein quantification and in turn the downstream analyses. (9−11) In contrast, performing imputation benefits statistical analyses as well as clustering methods as it increases data completeness.
Many review articles and benchmark studies investigate the performance of different imputation strategies applied to specific data types and acquisition techniques. (5,16,18,20,21) However, many authors claim that the results cannot be adapted to different data settings and a general recommendation is urgently needed. (5,12,22) This prompted us to develop DIMA, a general decision guide applicable to any proteomics data set, which reproduces the individual occurrence pattern of MVs and thereby suggests a suitable imputation method.
Although the mechanisms underlying MVs are diverse, the logistic regression model and the algorithm selection are able to capture this distribution without a priori knowledge. This is demonstrated by a simulation study where data with different MV percentages and MNAR/MCAR ratios are simulated and DIMA suggests one of the top three ranked algorithms compared to direct imputation assessment in all cases (Figure 3). Here, several algorithms show comparably good performance which indicates that, generally, selecting a well-performing imputation algorithm is sufficient. However, it is crucial to avoid a bad-performing imputation which may over- or underestimate the true data values (Supporting Information Figure S5) and therefore lead to analysis bias.
The selection and ranking of the imputation algorithms applied to 142 PRIDE data sets (Figure 2) reveal a similar algorithm selection as with NAguideR: (23) Examples of well-performing algorithms are ImpSeq, ImpSeqRob, and bpca, medium-performing methods are kNN, rf, cart, and norm, and SVD, mean/median, and MinDet/MinProb/QRILC seem to perform poorly. The best-performing imputation algorithms are the sequential algorithms of the R package rrcovNA which can thus serve as the best a priori choice for imputation. However, we recommend using a decision tool like DIMA or NAguideR to aid the decision progress for a suitable imputation algorithm.
Selecting an imputation algorithm highly depends on the downstream analysis steps. In most cases, evaluation based on the average distance to the truth is appropriate. For statistical testing, however, not only the average distance but also the variance should be maintained. Thus, besides the RMSE being an indicator for the accuracy of point estimates, further criteria such as RMSEt for the proximity of the t-statistics or the p-value pF of the F-statistics reflecting the variations in variances should be used to select an appropriate imputation algorithm.
For large data sets and/or time-saving purposes, one could consider omitting the calculation of imputation algorithms that are not expected to perform well based on prior knowledge e.g., algorithms, which were not suggested by DIMA for the PRIDE data sets. DIMA offers the possibility to specify the evaluated imputation approaches and provides a fast version including the nine algorithms that are most frequently recommended for the PRIDE data sets.
Methodically, DIMA is applicable to any data set in a matrix format independent of its acquisition. Specifically, data-dependent, data-independent, bottom-up, targeted, label-free, isotope labeling and more proteomics data sets can highly benefit from applying DIMA as it increases the quantification of peptides and proteins. In addition, DIMA could also be beneficial for data of other high-throughput techniques such as bulk or single-cell RNAseq data.
In summary, DIMA provides a realistic and data-dependent performance assessment and thereby suggests a suitable imputation algorithm. Its performance and effectiveness are demonstrated on simulated data sets as well as on 142 quantitative MS data sets.

Supporting Information

ARTICLE SECTIONS
Jump To

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.jproteome.1c00119.

  • Sigmoidal decrease of missing values for higher protein intensities (Figure S1); missing value distribution per sample and per protein (Figure S2); distribution of the estimated logistic regression coefficients (Figure S3); DIMA analysis at the peptide level (Figure S4); density plot of the imputed compared to the original data values (Figure S5); principal component analysis before and after imputation (Figure S6); DIMA Implementation (Figure S7); characteristics of the 30 applied imputation algorithms (Table S1); and characteristics of the PRIDE data sets assessed with DIMA (Table S2) (PDF)

Terms & Conditions

Most electronic Supporting Information files are available without a subscription to ACS Web Editions. Such files may be downloaded by article for research use (if there is a public use license linked to the relevant article, that license may permit other uses). Permission may be obtained from ACS for other uses through requests via the RightsLink permission system: http://pubs.acs.org/page/copyright/permissions.html.

Author Information

ARTICLE SECTIONS
Jump To
  • Corresponding Authors
    • Janine Egert - Institute of Medical Biometry and Statistics (IMBI), Institute of Medicine and Medical Center Freiburg, 79104 Freiburg im Breisgau, Germany;  Centre for Integrative Biological Signalling Studies (CIBSS), Albert-Ludwigs-Universität Freiburg, 79104 Freiburg, Germany;  Orcidhttps://orcid.org/0000-0002-2032-7081;  Email: [email protected]
    • Clemens Kreutz - Institute of Medical Biometry and Statistics (IMBI), Institute of Medicine and Medical Center Freiburg, 79104 Freiburg im Breisgau, Germany;  Signalling Research Centres BIOSS and CIBSS, Albert-Ludwigs-Universität Freiburg, 79104 Freiburg im Breisgau, Germany;  Center for Data Analysis and Modeling (FDM), Albert-Ludwigs-Universität Freiburg, 79104 Freiburg im Breisgau, Germany;  Email: [email protected]
  • Authors
    • Eva Brombacher - Institute of Medical Biometry and Statistics (IMBI), Institute of Medicine and Medical Center Freiburg, 79104 Freiburg im Breisgau, Germany;  Centre for Integrative Biological Signalling Studies (CIBSS), Albert-Ludwigs-Universität Freiburg, 79104 Freiburg, Germany;  Spemann Graduate School of Biology and Medicine (SGBM), Albert-Ludwigs-Universität Freiburg, 79104 Freiburg, Germany;  Faculty of Biology, Albert-Ludwigs-Universität Freiburg, 79104 Freiburg im Breisgau, Germany
    • Bettina Warscheid - Biochemistry and Functional Proteomics, Institute of Biology II, Faculty of Biology, Albert-Ludwigs-Universität Freiburg, 79104 Freiburg im Breisgau, Germany;  Signalling Research Centres BIOSS and CIBSS, Albert-Ludwigs-Universität Freiburg, 79104 Freiburg im Breisgau, Germany;  Orcidhttps://orcid.org/0000-0001-5096-1975
  • Notes
    The authors declare no competing financial interest.

Acknowledgments

ARTICLE SECTIONS
Jump To

This work was supported by the Federal Ministry of Education and Research of Germany [EA:Sys,FKZ031L0080 to J.E. and C.K.]; the Deutsche Forschungsgemeinschaft (German Research Foundation) [CIBSS-EXC-2189-2100249960-390939984 to E.B., B.W., and C.K., Project-ID 403222702278002225/SFB 1381 to B.W., FOR 2743 to B.W., TRR 130 to B.W.], and the European Research Council H2020 [648235 to B.W., Marie Sklodowska Curie grant 812968 to B.W.]. The authors acknowledge support from the state of Baden-Württemberg through bwHPC and the Deutsche Forschungsgemeinschaft through grant INST 35/1134-1 FUGG. The authors gratefully thank Lena Reimann, Wignand Mühlhäuser, and Friedel Drepper for fruitful discussions on the topic.

References

ARTICLE SECTIONS
Jump To

This article references 38 other publications.

  1. 1
    McGurk, K. A. The use of missing values in proteomic data-independent acquisition mass spectrometry to enable disease activity discrimination. Bioinformatics 2020, 36, 2217– 2223,  DOI: 10.1093/bioinformatics/btz898
    [Crossref], [PubMed], [CAS], Google Scholar
    1
    The use of missing values in proteomic data-independent acquisition mass spectrometry to enable disease activity discrimination
    McGurk, Kathryn A.; Dagliati, Arianna; Chiasserini, Davide; Lee, Dave; Plant, Darren; Baricevic-Jones, Ivona; Kelsall, Janet; Eineman, Rachael; Reed, Rachel; Geary, Bethany; Unwin, Richard D.; Nicolaou, Anna; Keavney, Bernard D.; Barton, Anne; Whetton, Anthony D.; Geifman, Nophar
    Bioinformatics (2020), 36 (7), 2217-2223CODEN: BOINFP; ISSN:1367-4811. (Oxford University Press)
    Data-independent acquisition mass spectrometry allows for comprehensive peptide detection and relative quantification than std. data-dependent approaches. While less prone to missing values, these still exist. Current approaches for handling the so-called missingness have challenges. We hypothesized that non-random missingness is a useful biol. measure and demonstrate the importance of analyzing missingness for proteomic discovery within a longitudinal study of disease activity. The magnitude of missingness did not correlate with mean peptide concn. The magnitude of missingness for each protein strongly correlated between collection time points (baseline, 3 mo, 6 mo; R = 0.95-0.97, confidence interval = 0.94-0.97) indicating little time-dependent effect. This allowed for the identification of proteins with outlier levels of missingness that differentiate between the patient groups characterized by different patterns of disease activity. The assocn. of these proteins with disease activity was confirmed by machine learning techniques. Our novel approach complements analyses on complete observations and other missing value strategies in biomarker prediction of disease activity.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3cXitlOmu7bI&md5=d2a3cd6644ada2086f6aaa6722145f68
  2. 2
    Poulos, R. C. Strategies to enable large-scale proteomics for reproducible research. Nat. Commun. 2020, 11, 3793  DOI: 10.1038/s41467-020-17641-3
    [Crossref], [PubMed], [CAS], Google Scholar
    2
    Strategies to enable large-scale proteomics for reproducible research
    Poulos, Rebecca C.; Hains, Peter G.; Shah, Rohan; Lucas, Natasha; Xavier, Dylan; Manda, Srikanth S.; Anees, Asim; Koh, Jennifer M. S.; Mahboob, Sadia; Wittman, Max; Williams, Steven G.; Sykes, Erin K.; Hecker, Michael; Dausmann, Michael; Wouters, Merridee A.; Ashman, Keith; Yang, Jean; Wild, Peter J.; deFazio, Anna; Balleine, Rosemary L.; Tully, Brett; Aebersold, Ruedi; Speed, Terence P.; Liu, Yansheng; Reddel, Roger R.; Robinson, Phillip J.; Zhong, Qing
    Nature Communications (2020), 11 (1), 3793CODEN: NCAOBW; ISSN:2041-1723. (Nature Research)
    Abstr.: Reproducible research is the bedrock of exptl. science. To enable the deployment of large-scale proteomics, we assess the reproducibility of mass spectrometry (MS) over time and across instruments and develop computational methods for improving quant. accuracy. We perform 1560 data independent acquisition (DIA)-MS runs of eight samples contg. known proportions of ovarian and prostate cancer tissue and yeast, or control HEK293T cells. Replicates are run on six mass spectrometers operating continuously with varying maintenance schedules over four months, interspersed with ∼5000 other runs. We utilize neg. controls and replicates to remove unwanted variation and enhance biol. signal, outperforming existing methods. We also design a method for reducing missing values. Integrating these computational modules into a pipeline (ProNorM), we mitigate variation among instruments over time and accurately predict tissue proportions. We demonstrate how to improve the quant. anal. of large-scale DIA-MS data, providing a pathway toward clin. proteomics.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3cXhsFajsb7M&md5=848bd6df41e1fc16efef3d87f76de510
  3. 3
    Brenes, A.; Hukelmann, J.; Bensaddek, D.; Lamond, A. I. Multibatch TMT Reveals False Positives, Batch Effects and Missing Values. Mol. Cell. Proteomics 2019, 18, 1967– 1980,  DOI: 10.1074/mcp.RA119.001472
    [Crossref], [PubMed], [CAS], Google Scholar
    3
    Multibatch TMT reveals false positives, batch effects and missing values
    Brenes, Alejandro; Hukelmann, Jens; Bensaddek, Dalila; Lamond, Angus I.
    Molecular & Cellular Proteomics (2019), 18 (10), 1967-1980CODEN: MCPOBS; ISSN:1535-9484. (American Society for Biochemistry and Molecular Biology)
    Multiplexing strategies for large-scale proteomic analyses have become increasingly prevalent, tandem mass tags (TMT) in particular. Here we used a large iPSC proteomic expt. with twenty-four 10-plex TMT batches to evaluate the effect of integrating multiple TMT batches within a single anal. We identified a significant inflation rate of protein missing values as multiple batches are integrated and show that this pattern is aggravated at the peptide level. We also show that without normalization strategies to address the batch effects, the high precision of quantitation within a single multiplexed TMT batch is not reproduced when data from multiple TMT batches are integrated. Further, the incidence of false positives was studied by using Y chromosome peptides as an internal control. The iPSC lines quantified in this data set were derived from both male and female donors, hence the peptides mapped to the Y chromosome should be absent from female lines. Nonetheless, these Y chromosome-specific peptides were consistently detected in the female channels of all TMT batches. We then used the same Y chromosome specific peptides to quantify the level of ion coisolation as well as the effect of primary and secondary reporter ion interference. These results were used to propose solns. to mitigate the limitations of multi-batch TMT analyses. We confirm that including a common ref. line in every batch increases precision by facilitating normalization across the batches and we propose exptl. designs that minimize the effect of cross population reporter ion interference.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC1MXitlSjtb3J&md5=1321490bdeb894c40aa34eff5d12546b
  4. 4
    Wei, R.; Wang, J.; Su, M.; Jia, E.; Chen, S.; Chen, T.; Ni, Y. Missing Value Imputation Approach for Mass Spectrometry-based Metabolomics Data. Sci. Rep. 2018, 8, 663  DOI: 10.1038/s41598-017-19120-0
    [Crossref], [PubMed], [CAS], Google Scholar
    4
    Missing Value Imputation Approach for Mass Spectrometry-based Metabolomics Data
    Wei Runmin; Wang Jingye; Su Mingming; Chen Shaoqiu; Ni Yan; Wei Runmin; Chen Shaoqiu; Su Mingming; Jia Erik; Chen Tianlu
    Scientific reports (2018), 8 (1), 663 ISSN:.
    Missing values exist widely in mass-spectrometry (MS) based metabolomics data. Various methods have been applied for handling missing values, but the selection can significantly affect following data analyses. Typically, there are three types of missing values, missing not at random (MNAR), missing at random (MAR), and missing completely at random (MCAR). Our study comprehensively compared eight imputation methods (zero, half minimum (HM), mean, median, random forest (RF), singular value decomposition (SVD), k-nearest neighbors (kNN), and quantile regression imputation of left-censored data (QRILC)) for different types of missing values using four metabolomics datasets. Normalized root mean squared error (NRMSE) and NRMSE-based sum of ranks (SOR) were applied to evaluate imputation accuracy. Principal component analysis (PCA)/partial least squares (PLS)-Procrustes analysis were used to evaluate the overall sample distribution. Student's t-test followed by correlation analysis was conducted to evaluate the effects on univariate statistics. Our findings demonstrated that RF performed the best for MCAR/MAR and QRILC was the favored one for left-censored MNAR. Finally, we proposed a comprehensive strategy and developed a public-accessible web-tool for the application of missing value imputation in metabolomics ( https://metabolomics.cc.hawaii.edu/software/MetImp/ ).
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A280%3ADC%252BC1MvgtlyrsQ%253D%253D&md5=f161109ca8155e59a728dd2761e994f6
  5. 5
    Webb-Robertson, B.-J. M.; Wiberg, H. K.; Matzke, M. M.; Brown, J. N.; Wang, J.; McDermott, J. E.; Smith, R. D.; Rodland, K. D.; Metz, T. O.; Pounds, J. G.; Waters, K. M. Reviewand Evaluationand and Discussion of the Challenges of Missing Value Imputation for Mass Spectrometry-Based Label-Free Global Proteomics. J. Proteome Res. 2015, 14, 1993– 2001,  DOI: 10.1021/pr501138h
    [ACS Full Text ACS Full Text], [CAS], Google Scholar
    5
    Review, Evaluation, and Discussion of the Challenges of Missing Value Imputation for Mass Spectrometry-Based Label-Free Global Proteomics
    Webb-Robertson, Bobbie-Jo M.; Wiberg, Holli K.; Matzke, Melissa M.; Brown, Joseph N.; Wang, Jing; McDermott, Jason E.; Smith, Richard D.; Rodland, Karin D.; Metz, Thomas O.; Pounds, Joel G.; Waters, Katrina M.
    Journal of Proteome Research (2015), 14 (5), 1993-2001CODEN: JPROBS; ISSN:1535-3893. (American Chemical Society)
    A review. In this review, we apply selected imputation strategies to label-free liq. chromatog.-mass spectrometry (LC-MS) proteomics datasets to evaluate the accuracy with respect to metrics of variance and classification. We evaluate several commonly used imputation approaches for individual merits and discuss the caveats of each approach with respect to the example LC-MS proteomics data. In general, local similarity-based approaches, such as the regularized expectation maximization and least-squares adaptive algorithms, yield the best overall performances with respect to metrics of accuracy and robustness. However, no single algorithm consistently outperforms the remaining approaches, and in some cases, performing classification without imputation sometimes yielded the most accurate classification. Thus, because of the complex mechanisms of missing data in proteomics, which also vary from peptide to protein, no individual method is a single soln. for imputation. On the basis of the observations in this review, the goal for imputation in the field of computational proteomics should be to develop new approaches that work generically for this data type and new strategies to guide users in the selection of the best imputation for their dataset and anal. objectives.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2MXmtV2ltL0%253D&md5=218ce277ccefd57aab2ad11987b943c6
  6. 6
    Lazar, C.; Laurent, G.; Myriam, F.; Christophe, B.; Thomas, B. Accounting for the Multiple Natures of Missing Values in Label-Free Quantitative Proteomics Data Sets to Compare Imputation Strategies. J. Proteome Res. 2016, 15, 1116– 1125,  DOI: 10.1021/acs.jproteome.5b00981
    [ACS Full Text ACS Full Text], [CAS], Google Scholar
    6
    Accounting for the Multiple Natures of Missing Values in Label-Free Quantitative Proteomics Data Sets to Compare Imputation Strategies
    Lazar, Cosmin; Gatto, Laurent; Ferro, Myriam; Bruley, Christophe; Burger, Thomas
    Journal of Proteome Research (2016), 15 (4), 1116-1125CODEN: JPROBS; ISSN:1535-3893. (American Chemical Society)
    Missing values are a genuine issue in label-free quant. proteomics. Recent works have surveyed the different statistical methods to conduct imputation and have compared them on real or simulated data sets and recommended a list of missing value imputation methods for proteomics application. Although insightful, these comparisons do not account for two important facts: (i) depending on the proteomics data set, the missingness mechanism may be of different natures and (ii) each imputation method is devoted to a specific type of missingness mechanism. As a result, we believe that the question at stake is not to find the most accurate imputation method in general but instead the most appropriate one. We describe a series of comparisons that support our views: For instance, we show that a supposedly "under-performing" method (i.e., giving baseline av. results), if applied at the "appropriate" time in the data-processing pipeline (before or after peptide aggregation) on a data set with the "appropriate" nature of missing values, can outperform a blindly applied, supposedly "better-performing" method (i.e., the ref. method from the state-of-the-art). This leads us to formulate few practical guidelines regarding the choice and the application of an imputation method in a proteomics context.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC28XivFOntbw%253D&md5=edecca833d3f9183fb1fe3e9ff14e8af
  7. 7
    Rubin, D. B. Inference and missing data. Biometrika 1976, 63, 581– 592,  DOI: 10.1093/biomet/63.3.581
    [Crossref], Google Scholar
    There is no corresponding record for this reference.
  8. 8
    Karpievitch, Y. V.; Dabney, A. R.; Smith, R. D. Normalization and missing value imputation for label-free LC-MS analysis. BMC Bioinf. 2012, 13, S5  DOI: 10.1186/1471-2105-13-S16-S5
    [Crossref], [PubMed], [CAS], Google Scholar
    8
    Normalization and missing value imputation for label-free LC-MS analysis
    Karpievitch, Yuliya V.; Dabney, Alan R.; Smith, Richard D.
    BMC Bioinformatics (2012), 13 (Suppl. 16), S5CODEN: BBMIC4; ISSN:1471-2105. (BioMed Central Ltd.)
    Shotgun proteomic data are affected by a variety of known and unknown systematic biases as well as high proportions of missing values. Typically, normalization is performed in an attempt to remove systematic biases from the data before statistical inference, sometimes followed by missing value imputation to obtain a complete matrix of intensities. Here we discuss several approaches to normalization and dealing with missing values, some initially developed for microarray data and some developed specifically for mass spectrometry-based data.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC38XhvVyis7bN&md5=f1dbc6f041dba77202a072e3e0f4a959
  9. 9
    Välikangas, T.; Suomi, T.; Elo, L. L. A comprehensive evaluation of popular proteomics software workflows for label-free proteome quantification and imputation. Briefings Bioinf. 2017, 19, 1344– 1355,  DOI: 10.1093/bib/bbx054
    [Crossref], Google Scholar
    There is no corresponding record for this reference.
  10. 10
    Wang, J.; Li, L.; Chen, T.; Ma, J.; Zhu, Y.; Zhuang, J.; Chang, C. In-depth method assessments of differentially expressed protein detection for shotgun proteomics data with missing values. Sci. Rep. 2017, 7, 3367  DOI: 10.1038/s41598-017-03650-8
    [Crossref], [PubMed], [CAS], Google Scholar
    10
    In-depth method assessments of differentially expressed protein detection for shotgun proteomics data with missing values
    Wang Jinxia; Li Liwei; Chen Tao; Ma Jie; Zhu Yunping; Chang Cheng; Wang Jinxia; Zhuang Jujuan; Wang Jinxia
    Scientific reports (2017), 7 (1), 3367 ISSN:.
    Considering as one of the major goals in quantitative proteomics, detection of the differentially expressed proteins (DEPs) plays an important role in biomarker selection and clinical diagnostics. There have been plenty of algorithms and tools focusing on DEP detection in proteomics research. However, due to the different application scopes of these methods, and various kinds of experiment designs, it is not very apparent about the best choice for large-scale proteomics data analyses. Moreover, given the fact that proteomics data usually contain high percentage of missing values (MVs), but few replicates, a systematic evaluation of the DEP detection methods combined with the MV imputation methods is essential and urgent. Here, we analyzed a total of four representative imputation methods and five DEP methods on different experimental and simulated datasets. The results showed that (i) MV imputation could not always improve the performances of DEP detection methods and the imputation effects differed in the missing value percentages; (ii) the DEP detection methods had different statistical powers affected by the percentage of MVs. Two statistical methods (i.e. the empirical Bayesian random censoring threshold model, and the significance analysis of microarray) performed better than the other evaluated methods in terms of accuracy and sensitivity.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A280%3ADC%252BC1cnnvF2iug%253D%253D&md5=038103fc16b08f4e3b802f37c1172b39
  11. 11
    Janssen, K. J.; Donders, A. R. T.; Harrell, F. E.; Vergouwe, Y.; Chen, Q.; Grobbee, D. E.; Moons, K. G. Missing covariate data in medical research: To impute is better than to ignore. J. Clin. Epidemiol. 2010, 63, 721– 727,  DOI: 10.1016/j.jclinepi.2009.12.008
    [Crossref], [PubMed], [CAS], Google Scholar
    11
    Missing covariate data in medical research: to impute is better than to ignore
    Janssen Kristel J M; Donders A Rogier T; Harrell Frank E Jr; Vergouwe Yvonne; Chen Qingxia; Grobbee Diederick E; Moons Karel G M
    Journal of clinical epidemiology (2010), 63 (7), 721-7 ISSN:.
    OBJECTIVE: We compared popular methods to handle missing data with multiple imputation (a more sophisticated method that preserves data). STUDY DESIGN AND SETTING: We used data of 804 patients with a suspicion of deep venous thrombosis (DVT). We studied three covariates to predict the presence of DVT: d-dimer level, difference in calf circumference, and history of leg trauma. We introduced missing values (missing at random) ranging from 10% to 90%. The risk of DVT was modeled with logistic regression for the three methods, that is, complete case analysis, exclusion of d-dimer level from the model, and multiple imputation. RESULTS: Multiple imputation showed less bias in the regression coefficients of the three variables and more accurate coverage of the corresponding 90% confidence intervals than complete case analysis and dropping d-dimer level from the analysis. Multiple imputation showed unbiased estimates of the area under the receiver operating characteristic curve (0.88) compared with complete case analysis (0.77) and when the variable with missing values was dropped (0.65). CONCLUSION: As this study shows that simple methods to deal with missing data can lead to seriously misleading results, we advise to consider multiple imputation. The purpose of multiple imputation is not to create data, but to prevent the exclusion of observed data.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A280%3ADC%252BC3czmtV2gtQ%253D%253D&md5=bab541926c308b8b5c901d42982dae27
  12. 12
    Brock, G. N.; Shaffer, J. R.; Blakesley, R. E.; Lotz, M. J.; Tseng, G. C. Which missing value imputation method to use in expression profiles: a comparative study and two selection schemes. BMC Bioinf. 2008, 9, 12  DOI: 10.1186/1471-2105-9-12
    [Crossref], [PubMed], [CAS], Google Scholar
    12
    Which missing value imputation method to use in expression profiles: a comparative study and two selection schemes
    Brock Guy N; Shaffer John R; Blakesley Richard E; Lotz Meredith J; Tseng George C
    BMC bioinformatics (2008), 9 (), 12 ISSN:.
    BACKGROUND: Gene expression data frequently contain missing values, however, most down-stream analyses for microarray experiments require complete data. In the literature many methods have been proposed to estimate missing values via information of the correlation patterns within the gene expression matrix. Each method has its own advantages, but the specific conditions for which each method is preferred remains largely unclear. In this report we describe an extensive evaluation of eight current imputation methods on multiple types of microarray experiments, including time series, multiple exposures, and multiple exposures x time series data. We then introduce two complementary selection schemes for determining the most appropriate imputation method for any given data set. RESULTS: We found that the optimal imputation algorithms (LSA, LLS, and BPCA) are all highly competitive with each other, and that no method is uniformly superior in all the data sets we examined. The success of each method can also depend on the underlying "complexity" of the expression data, where we take complexity to indicate the difficulty in mapping the gene expression matrix to a lower-dimensional subspace. We developed an entropy measure to quantify the complexity of expression matrixes and found that, by incorporating this information, the entropy-based selection (EBS) scheme is useful for selecting an appropriate imputation algorithm. We further propose a simulation-based self-training selection (STS) scheme. This technique has been used previously for microarray data imputation, but for different purposes. The scheme selects the optimal or near-optimal method with high accuracy but at an increased computational cost. CONCLUSION: Our findings provide insight into the problem of which imputation method is optimal for a given data set. Three top-performing methods (LSA, LLS and BPCA) are competitive with each other. Global-based imputation methods (PLS, SVD, BPCA) performed better on mcroarray data with lower complexity, while neighbour-based methods (KNN, OLS, LSA, LLS) performed better in data with higher complexity. We also found that the EBS and STS schemes serve as complementary and effective tools for selecting the optimal imputation algorithm.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A280%3ADC%252BD1c7jsVyhug%253D%253D&md5=550ac2936960ddcd03d9ca2586c4281d
  13. 13
    To, K. T.; Fry, R. C.; Reif, D. M. Characterizing the effects of missing data and evaluating imputation methods for chemical prioritization applications using ToxPi. BioData Min. 2018, 11, 10  DOI: 10.1186/s13040-018-0169-5
    [Crossref], [PubMed], [CAS], Google Scholar
    13
    Characterizing the effects of missing data and evaluating imputation methods for chemical prioritization applications using ToxPi
    To Kimberly T; Reif David M; Fry Rebecca C; Reif David M; Reif David M
    BioData mining (2018), 11 (), 10 ISSN:1756-0381.
    BACKGROUND: The Toxicological Priority Index (ToxPi) is a method for prioritization and profiling of chemicals that integrates data from diverse sources. However, individual data sources ("assays"), such as in vitro bioassays or in vivo study endpoints, often feature sections of missing data, wherein subsets of chemicals have not been tested in all assays. In order to investigate the effects of missing data and recommend solutions, we designed simulation studies around high-throughput screening data generated by the ToxCast and Tox21 programs on chemicals highlighted by the Agency for Toxic Substances and Disease Registry's (ATSDR) Substance Priority List (SPL), which helps prioritize environmental research and remediation resources. RESULTS: Our simulations explored a wide range of scenarios concerning data (0-80% assay data missing per chemical), modeling (ToxPi models containing from 160-700 different assays), and imputation method (k-Nearest-Neighbor, Max, Mean, Min, Binomial, Local Least Squares, and Singular Value Decomposition). We find that most imputation methods result in significant changes to ToxPi score, except for datasets with a small number of assays. If we consider rank change conditional on these significant changes to ToxPi score, we find that ranks of chemicals in the minimum value imputation, SVD imputation, and kNN imputation sets are more sensitive to the score changes. CONCLUSIONS: We found that the choice of imputation strategy exerted significant influence over both scores and associated ranks, and the most sensitive scenarios were those involving fewer assays plus higher proportions of missing data. By characterizing the effects of missing data and the relative benefit of imputation approaches across real-world data scenarios, we can augment confidence in the robustness of decisions regarding the health and ecological effects of environmental chemicals.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A280%3ADC%252BB3c%252FhslOksQ%253D%253D&md5=5ea3228f441ab27190b2ada91f5059e6
  14. 14
    Poyatos, R.; Sus, O.; Badiella, L.; Mencuccini, M.; Martinez-Vilalta, J. Gap-filling a spatially explicit plant trait database: comparing imputation methods and different levels of environmental information. Biogeosciences 2018, 15, 2601– 2617,  DOI: 10.5194/bg-15-2601-2018
    [Crossref], Google Scholar
    There is no corresponding record for this reference.
  15. 15
    Lenz, M.; Schulz, A.; Koeck, T.; Rapp, S.; Nagler, M.; Sauer, M.; Eggebrecht, L.; Cate, V. T.; Panova-Noeva, M.; Prochaska, J. H.; Lackner, K. J.; Münzel, T.; Leineweber, K.; Wild, P. S.; Andrade-Navarro, M. A. Missing value imputation in proximity extension assay-based targeted proteomics data. PLoS One 2020, 15, e0243487  DOI: 10.1371/journal.pone.0243487
    [Crossref], [PubMed], [CAS], Google Scholar
    15
    Missing value imputation in proximity extension assay-based targeted proteomics data
    Lenz, Michael; Schulz, Andreas; Koeck, Thomas; Rapp, Steffen; Nagler, Markus; Sauer, Madeleine; Eggebrecht, Lisa; Ten Cate, Vincent; Panova-Noeva, Marina; Prochaska, Juergen H.; Lackner, Karl J.; Muenzel, Thomas; Leineweber, Kirsten; Wild, Philipp S.; Andrade-Navarro, Miguel A.
    PLoS One (2020), 15 (12), e0243487CODEN: POLNCL; ISSN:1932-6203. (Public Library of Science)
    Targeted proteomics utilizing antibody-based proximity extension assays provides sensitive and highly specific quantifications of plasma protein levels. Multivariate anal. of this data is hampered by frequent missing values (random or left censored), calling for imputation approaches. While appropriate missing-value imputation methods exist, benchmarks of their performance in targeted proteomics data are lacking. Here, we assessed the performance of two methods for imputation of values missing completely at random, the previously top-benchmarked 'missForest' and the recently published 'GSimp' method. Evaluation was accomplished by comparing imputed with remeasured relative concns. of 91 inflammation related circulating proteins in 86 samples from a cohort of 645 patients with venous thromboembolism. The median Pearson correlation between imputed and remeasured protein expression values was 69.0% for missForest and 71.6% for GSimp (p = 5.8e-4). Imputation with missForest resulted in stronger redn. of variance compared to GSimp (median relative variance of 25.3% vs. 68.6%, p = 2.4e-16) and undesired larger bias in downstream anal. Irresp. of the imputation method used, the 91 imputed proteins revealed large variations in imputation accuracy, driven by differences in signal to noise ratio and information overlap between proteins. In summary, GSimp outperformed missForest, while both methods show good overall imputation accuracy with large variations between proteins.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3cXis1GgsrrI&md5=a13ae3b2e3fc861f525e13f71405736d
  16. 16
    Bramer, L. M.; Irvahn, J.; Piehowski, P. D.; Rodland, K. D.; Webb-Robertson, B.-J. M. A Review of Imputation Strategies for Isobaric Labeling-Based Shotgun Proteomics. J. Proteome Res. 2021, 20, 1– 13,  DOI: 10.1021/acs.jproteome.0c00123
    [ACS Full Text ACS Full Text], [CAS], Google Scholar
    16
    A Review of Imputation Strategies for Isobaric Labeling-Based Shotgun Proteomics
    Bramer, Lisa M.; Irvahn, Jan; Piehowski, Paul D.; Rodland, Karin D.; Webb-Robertson, Bobbie-Jo M.
    Journal of Proteome Research (2021), 20 (1), 1-13CODEN: JPROBS; ISSN:1535-3893. (American Chemical Society)
    A review. The throughput efficiency and increased depth of coverage provided by isobaric-labeled proteomics measurements have led to increased usage of these techniques. However, the structure of missing data is different than unlabeled studies, which prompts the need for this review to compare the efficacy of nine imputation methods on large isobaric-labeled proteomics data sets to guide researchers on the appropriateness of various imputation methods. Imputation methods were evaluated by accuracy, statistical hypothesis test inference, and run time. In general, expectation maximization and random forest imputation methods yielded the best performance, and const.-based methods consistently performed poorly across all data set sizes and percentages of missing values. For data sets with small sample sizes and higher percentages of missing data, results indicate that statistical inference with no imputation may be preferable. On the basis of the findings in this review, there are core imputation methods that perform better for isobaric-labeled proteomics data, but great care and consideration as to whether imputation is the optimal strategy should be given for data sets comprised of a small no. of samples.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3cXhvVerur3N&md5=6de86677050965321a2b7a05314cb1f7
  17. 17
    de Souto, M. C. P.; Jaskowiak, P. A.; Costa, I. G. Impact of missing data imputation methods on gene expression clustering and classification. BMC Bioinf. 2015, 16, 64  DOI: 10.1186/s12859-015-0494-3
    [Crossref], [PubMed], [CAS], Google Scholar
    17
    Impact of missing data imputation methods on gene expression clustering and classification
    de Souto Marcilio C P; Jaskowiak Pablo A; Costa Ivan G; Costa Ivan G
    BMC bioinformatics (2015), 16 (), 64 ISSN:.
    BACKGROUND: Several missing value imputation methods for gene expression data have been proposed in the literature. In the past few years, researchers have been putting a great deal of effort into presenting systematic evaluations of the different imputation algorithms. Initially, most algorithms were assessed with an emphasis on the accuracy of the imputation, using metrics such as the root mean squared error. However, it has become clear that the success of the estimation of the expression value should be evaluated in more practical terms as well. One can consider, for example, the ability of the method to preserve the significant genes in the dataset, or its discriminative/predictive power for classification/clustering purposes. RESULTS AND CONCLUSIONS: We performed a broad analysis of the impact of five well-known missing value imputation methods on three clustering and four classification methods, in the context of 12 cancer gene expression datasets. We employed a statistical framework, for the first time in this field, to assess whether different imputation methods improve the performance of the clustering/classification methods. Our results suggest that the imputation methods evaluated have a minor impact on the classification and downstream clustering analyses. Simple methods such as replacing the missing values by mean or the median values performed as well as more complex strategies. The datasets analyzed in this study are available at http://costalab.org/Imputation/ .
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A280%3ADC%252BC2Mjlslemsg%253D%253D&md5=210bc346aff2d28f775b2c1f2a4093b7
  18. 18
    Liu, M.; Dongre, A. Proper imputation of missing values in proteomics datasets for differential expression analysis. Briefings Bioinf. 2020, 0, bbaa112  DOI: 10.1093/bib/bbaa112
    [Crossref], Google Scholar
    There is no corresponding record for this reference.
  19. 19
    Rodwell, L.; Lee, K. J.; Romaniuk, H.; Carlin, J. B. Comparison of methods for imputing limited-range variables: a simulation study. BMC Med. Res. Methodol. 2014, 14, 57  DOI: 10.1186/1471-2288-14-57
    [Crossref], [PubMed], [CAS], Google Scholar
    19
    Comparison of methods for imputing limited-range variables: a simulation study
    Rodwell Laura; Lee Katherine J; Romaniuk Helena; Carlin John B
    BMC medical research methodology (2014), 14 (), 57 ISSN:.
    BACKGROUND: Multiple imputation (MI) was developed as a method to enable valid inferences to be obtained in the presence of missing data rather than to re-create the missing values. Within the applied setting, it remains unclear how important it is that imputed values should be plausible for individual observations. One variable type for which MI may lead to implausible values is a limited-range variable, where imputed values may fall outside the observable range. The aim of this work was to compare methods for imputing limited-range variables, with a focus on those that restrict the range of the imputed values. METHODS: Using data from a study of adolescent health, we consider three variables based on responses to the General Health Questionnaire (GHQ), a tool for detecting minor psychiatric illness. These variables, based on different scoring methods for the GHQ, resulted in three continuous distributions with mild, moderate and severe positive skewness. In an otherwise complete dataset, we set 33% of the GHQ observations to missing completely at random or missing at random; repeating this process to create 1000 datasets with incomplete data for each scenario.For each dataset, we imputed values on the raw scale and following a zero-skewness log transformation using: univariate regression with no rounding; post-imputation rounding; truncated normal regression; and predictive mean matching. We estimated the marginal mean of the GHQ and the association between the GHQ and a fully observed binary outcome, comparing the results with complete data statistics. RESULTS: Imputation with no rounding performed well when applied to data on the raw scale. Post-imputation rounding and imputation using truncated normal regression produced higher marginal means than the complete data estimate when data had a moderate or severe skew, and this was associated with under-coverage of the complete data estimate. Predictive mean matching also produced under-coverage of the complete data estimate. For the estimate of association, all methods produced similar estimates to the complete data. CONCLUSIONS: For data with a limited range, multiple imputation using techniques that restrict the range of imputed values can result in biased estimates for the marginal mean when data are highly skewed.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A280%3ADC%252BC2cnmvVKlsA%253D%253D&md5=67b590a65a7366914e38279d9ed1cb6a
  20. 20
    Kruttika, D.; Simion, K.; R, J. M.; J, P. S. A Simple Optimization Workflow to Enable Precise and Accurate Imputation of Missing Values in Proteomic Datasets. bioRxiv 2020, 1– 39,  DOI: 10.1101/2019.12.11.123456
    [Crossref], Google Scholar
    There is no corresponding record for this reference.
  21. 21
    Jin, L.; Bi, Y.; Hu, C.; Qu, J.; Shen, S.; Wang, X.; Tian, Y. A comparative study of evaluating missing value imputation methodsin label-free proteomics. Sci. Rep. 2021, 11, 1760  DOI: 10.1038/s41598-021-81279-4
    [Crossref], [PubMed], [CAS], Google Scholar
    21
    A comparative study of evaluating missing value imputation methods in label-free proteomics
    Jin, Liang; Bi, Yingtao; Hu, Chenqi; Qu, Jun; Shen, Shichen; Wang, Xue; Tian, Yu
    Scientific Reports (2021), 11 (1), 1760CODEN: SRCEC3; ISSN:2045-2322. (Nature Research)
    The presence of missing values (MVs) in label-free quant. proteomics greatly reduces the completeness of data. Imputation has been widely utilized to handle MVs, and selection of the proper method is crit. for the accuracy and reliability of imputation. Here we present a comparative study that evaluates the performance of seven popular imputation methods with a large-scale benchmark dataset and an immune cell dataset. Simulated MVs were incorporated into the complete part of each dataset with different combinations of MV rates and missing not at random (MNAR) rates. Normalized root mean square error (NRMSE) was applied to evaluate the accuracy of protein abundances and intergroup protein ratios after imputation. Detection of true positives (TPs) and false altered-protein discovery rate (FADR) between groups were also compared using the benchmark dataset. Furthermore, the accuracy of handling real MVs was assessed by comparing enriched pathways and signature genes of cell activation after imputing the immune cell dataset. We obsd. that the accuracy of imputation is primarily affected by the MNAR rate rather than the MV rate, and downstream anal. can be largely impacted by the selection of imputation methods. A random forest-based imputation method consistently outperformed other popular methods by achieving the lowest NRMSE, high amt. of TPs with the av. FADR < 5%, and the best detection of relevant pathways and signature genes, highlighting it as the most suitable method for label-free proteomics.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3MXhvFaqtrk%253D&md5=34e7040ccb9fb0dfe9b7e371dac1bcb1
  22. 22
    Audoux, J.; Salson, M.; Grosset, C. F.; Beaumeunier, S.; Holder, J.-M.; Commes, T.; Philippe, N. SimBA: A methodology and tools for evaluating the performance of RNA-Seq bioinformatic pipelines. BMC Bioinf. 2017, 18, 428  DOI: 10.1186/s12859-017-1831-5
    [Crossref], [PubMed], [CAS], Google Scholar
    22
    SimBA: a methodology and tools for evaluating the performance of RNA-Seq bioinformatic pipelines
    Audoux, Jerome; Salson, Mikael; Grosset, Christophe F.; Beaumeunier, Sacha; Holder, Jean-Marc; Commes, Therese; Philippe, Nicolas
    BMC Bioinformatics (2017), 18 (), 428/1-428/14CODEN: BBMIC4; ISSN:1471-2105. (BioMed Central Ltd.)
    The evolution of next-generation sequencing (NGS) technologies has led to increased focus on RNA-Seq. Many bioinformatic tools have been developed for RNA-Seq anal., each with unique performance characteristics and configuration parameters. Users face an increasingly complex task in understanding which bioinformatic tools are best for their specific needs and how they should be configured. In order to provide some answers to these questions, we investigate the performance of leading bioinformatic tools designed for RNA-Seq anal. and propose a methodol. for systematic evaluation and comparison of performance to help users make well informed choices. To evaluate RNA-Seq pipelines, we developed a suite of two benchmarking tools. SimCT generates simulated datasets that get as close as possible to specific real biol. conditions accompanied by the list of genomic incidents and mutations that have been inserted. BenchCT then compares the output of any bioinformatics pipeline that has been run against a SimCT dataset with the simulated genomic and transcriptional variations it contains to give an accurate performance evaluation in addressing specific biol. question. We used these tools to simulate a real-world genomic medicine questions involving the comparison of healthy and cancerous cells. Results revealed that performance in addressing a particular biol. context varied significantly depending on the choice of tools and settings used. We also found that by combining the output of certain pipelines, substantial performance improvements could be achieved. Our research emphasizes the importance of selecting and configuring bioinformatic tools for the specific biol. question being investigated to obtain optimal results. Pipeline designers, developers and users should include benchmarking in the context of their biol. question as part of their design and quality control process. Our SimBA suite of benchmarking tools provides a reliable basis for comparing the performance of RNA-Seq bioinformatics pipelines in addressing a specific biol. question. We would like to see the creation of a ref. corpus of data-sets that would allow accurate comparison between benchmarks performed by different groups and the publication of more benchmarks based on this public corpus.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC1cXitVOhtbrM&md5=024e4f119feb752a3458f3a643e71527
  23. 23
    Wang, S.; Li, W.; Hu, L.; Cheng, J.; Yang, H.; Liu, Y. NAguideR: performing and prioritizing missing value imputations for consistent bottom-up proteomic analyses. Nucleic Acids Res. 2020, 48, e83,  DOI: 10.1093/nar/gkaa498
    [Crossref], [PubMed], [CAS], Google Scholar
    23
    NAguideR: performing and prioritizing missing value imputations for consistent bottom-up proteomic analyses
    Wang, Shisheng; Li, Wenxue; Hu, Liqiang; Cheng, Jingqiu; Yang, Hao; Liu, Yansheng
    Nucleic Acids Research (2020), 48 (14), e83CODEN: NARHAD; ISSN:1362-4962. (Oxford University Press)
    Mass spectrometry (MS)-based quant. proteomics expts. frequently generate data with missing values, which may profoundly affect downstream analyses. A wide variety of imputation methods have been established to deal with the missing-value issue. To date, however, there is a scarcity of efficient, systematic, and easy-to-handle tools that are tailored for proteomics community. Herein, we developed a user-friendly and powerful stand-alone software, NAguideR, to enable implementation and evaluation of different missing value methods offered by 23 widely used missing-value imputation algorithms. NAguideR further evaluates data imputation results through classic computational criteria and, unprecedentedly, proteomic empirical criteria, such as quant. consistency between different charge-states of the same peptide, different peptides belonging to the same proteins, and individual proteins participating protein complexes and functional interactions. We applied NAguideR into three label-free proteomic datasets featuring peptide-level, protein-level, and phosphoproteomic variables resp., all generated by data independent acquisition mass spectrometry (DIA-MS) with substantial biol. replicates. The results indicate that NAguideR is able to discriminate the optimal imputation methods that are facilitating DIA-MS expts. over those sub-optimal and low-performance algorithms. NAguideR further provides downloadable tables and figures supporting flexible data anal. and interpretation.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3cXis1entrrE&md5=ff8b871b38806a95b7fb0220d5782daf
  24. 24
    Rose, M.; Duhamel, M.; Aboulouard, S.; Kobeissy, F.; Rhun, E. L.; Desmons, A.; Tierny, D.; Fournier, I.; Rodet, F.; Salzet, M. The Role of a Proprotein Convertase Inhibitor in Reactivation of Tumor-Associated Macrophages and Inhibition of Glioma Growth. Mol. Ther.--Oncolytics 2020, 17, 31– 46,  DOI: 10.1016/j.omto.2020.03.005
    [Crossref], [PubMed], [CAS], Google Scholar
    24
    The Role of a Proprotein Convertase Inhibitor in Reactivation of Tumor-Associated Macrophages and Inhibition of Glioma Growth
    Rose, Melanie; Duhamel, Marie; Aboulouard, Soulaimane; Kobeissy, Firas; Le Rhun, Emilie; Desmons, Annie; Tierny, Dominique; Fournier, Isabelle; Rodet, Franck; Salzet, Michel
    Molecular Therapy--Oncolytics (2020), 17 (), 31-46CODEN: MTOHDL; ISSN:2372-7705. (Elsevier Inc.)
    Tumors are characterized by the presence of malignant and non-malignant cells, such as immune cells including macrophages, which are preponderant. Macrophages impact the efficacy of chemotherapy and may lead to drug resistance. In this context and based on our previous work, we investigated the ability to reactivate macrophages by using a proprotein convertases inhibitor. Proprotein convertases process immature proteins into functional proteins, with several of them having a role in immune cell activation and tumorigenesis. Macrophages were treated with a peptidomimetic inhibitor targeting furin, PC1/3, PC4, PACE4, and PC5/6. Their anti-glioma activity was analyzed by mass spectrometry-based proteomics and viability assays in 2D and 3D in vitro cultures. Comparison with temozolomide, the drug used for glioma therapy, established that the inhibitor was more efficient for the redn. of cancer cell d. The inhibitor was also able to reactivate macrophages through the secretion of several immune factors with antitumor properties. Moreover, two proteins considered as good glioma patient survival indicators were also identified in 3D cultures treated with the inhibitor. Finally, we established that the proprotein convertases inhibitor has a dual role as an anti-glioma drug and anti-tumoral macrophage reactivation drug. This strategy could be used together with chemotherapy to increase therapy efficacy in glioma.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3cXht1GntbnL&md5=8198a0c313c461e11bafb3372d34b112
  25. 25
    Cox, J.; Mann, M. MaxQuant enables high peptide identification rates and individualized p.p.b.-range mass accuracies and proteome-wide protein quantification. Nat. Biotechnol. 2008, 26, 1367– 1372,  DOI: 10.1038/nbt.1511
    [Crossref], [PubMed], [CAS], Google Scholar
    25
    MaxQuant enables high peptide identification rates, individualized p.p.b.-range mass accuracies and proteome-wide protein quantification
    Cox, Juergen; Mann, Matthias
    Nature Biotechnology (2008), 26 (12), 1367-1372CODEN: NABIF9; ISSN:1087-0156. (Nature Publishing Group)
    Efficient anal. of very large amts. of raw data for peptide identification and protein quantification is a principal challenge in mass spectrometry (MS)-based proteomics. Here we describe MaxQuant, an integrated suite of algorithms specifically developed for high-resoln., quant. MS data. Using correlation anal. and graph theory, MaxQuant detects peaks, isotope clusters and stable amino acid isotope-labeled (SILAC) peptide pairs as three-dimensional objects in m/z, elution time and signal intensity space. By integrating multiple mass measurements and correcting for linear and nonlinear mass offsets, we achieve mass accuracy in the p.p.b. range, a sixfold increase over std. techniques. We increase the proportion of identified fragmentation spectra to 73% for SILAC peptide pairs via unambiguous assignment of isotope and missed-cleavage state and individual mass precision. MaxQuant automatically quantifies several hundred thousand peptides per SILAC-proteome expt. and allows statistically robust identification and quantification of >4000 proteins in mammalian cell lysates.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BD1cXhsVWjtLzJ&md5=675d31ca84e3a7e4fb9bdd601d8075ea
  26. 26
    O’Brien, J. J.; Gunawardena, H. P.; Paulo, J. A.; Chen, X.; Ibrahim, J. G.; Gygi, S. P.; Qaqish, B. F. The effects of nonignorable missing data on label-free mass spectrometry proteomics experiments. Ann. Appl. Stat. 2018, 12, 2075– 2095,  DOI: 10.1214/18-AOAS1144
    [Crossref], [PubMed], [CAS], Google Scholar
    26
    The effects of nonignorable missing data on label-free mass spectrometry proteomics experiments
    O'Brien Jonathon J; Gunawardena Harsha P; Paulo Joao A; Chen Xian; Ibrahim Joseph G; Gygi Steven P; Qaqish Bahjat F
    The annals of applied statistics (2018), 12 (4), 2075-2095 ISSN:1932-6157.
    An idealized version of a label-free discovery mass spectrometry proteomics experiment would provide absolute abundance measurements for a whole proteome, across varying conditions. Unfortunately, this ideal is not realized. Measurements are made on peptides requiring an inferential step to obtain protein level estimates. The inference is complicated by experimental factors that necessitate relative abundance estimation and result in widespread non-ignorable missing data. Relative abundance on the log scale takes the form of parameter contrasts. In a complete-case analysis, contrast estimates may be biased by missing data and a substantial amount of useful information will often go unused. To avoid problems with missing data, many analysts have turned to single imputation solutions. Unfortunately, these methods often create further difficulties by hiding inestimable contrasts, preventing the recovery of interblock information and failing to account for imputation uncertainty. To mitigate many of the problems caused by missing values, we propose the use of a Bayesian selection model. Our model is tested on simulated data, real data with simulated missing values, and on a ground truth dilution experiment where all of the true relative changes are known. The analysis suggests that our model, compared with various imputation strategies and complete-case analyses, can increase accuracy and provide substantial improvements to interval coverage.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A280%3ADC%252BB3crksFeitA%253D%253D&md5=e7433543cf49d814974145824b08c735
  27. 27
    Marquardt, D. W. Comment - You should standardize the predictor variables in your regression models. J. Am. Stat. Assoc. 1980, 75, 87– 91, 10.1080/01621459.1980.10477430
    Google Scholar
    There is no corresponding record for this reference.
  28. 28
    Menard, S. Standards for standardized logistic regression coefficients. Soc. Forces 2011, 89, 1409– 1428,  DOI: 10.1093/sf/89.4.1409
    [Crossref], Google Scholar
    There is no corresponding record for this reference.
  29. 29
    Kreutz, C. New Concepts for Evaluating the Performance of Computational Methods. IFAC-PapersOnLine 2016, 49, 63– 70,  DOI: 10.1016/j.ifacol.2016.12.104
    [Crossref], Google Scholar
    There is no corresponding record for this reference.
  30. 30
    MATLAB. 9.8.0.1538580 (R2020a); The MathWorks Inc.: Natickand Massachusetts, 2020.
    Google Scholar
    There is no corresponding record for this reference.
  31. 31
    R Core Team. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing. Vienna, Austria, 2021.
    Google Scholar
    There is no corresponding record for this reference.
  32. 32
    Stekhoven, D. J.; Buhlmann, P. MissForest-non-parametric missing value imputation for mixed-type data. Bioinformatics 2012, 28, 112– 118,  DOI: 10.1093/bioinformatics/btr597
    [Crossref], [PubMed], [CAS], Google Scholar
    32
    MissForest-non-parametric missing value imputation for mixed-type data
    Stekhoven, Daniel J.; Buehlmann, Peter
    Bioinformatics (2012), 28 (1), 112-118CODEN: BOINFP; ISSN:1367-4803. (Oxford University Press)
    Modern data acquisition based on high-throughput technol. is often facing the problem of missing data. Algorithms commonly used in the anal. of such large-scale data often depend on a complete set. Missing value imputation offers a soln. to this problem. However, the majority of available imputation methods are restricted to one type of variable only: continuous or categorical. For mixed-type data, the different types are usually handled sep. Therefore, these methods ignore possible relations between variable types. We propose a non-parametric method which can cope with different types of variables simultaneously. Results: We compare several state of the art methods for the imputation of missing values. We propose and evaluate an iterative imputation method (missForest) based on a random forest. By averaging over many unpruned classification or regression trees, random forest intrinsically constitutes a multiple imputation scheme. Using the built-in out-of-bag error ests. of random forest, we are able to est. the imputation error without the need of a test set. Evaluation is performed on multiple datasets coming from a diverse selection of biol. fields with artificially introduced missing values ranging from 10% to 30%. We show that missForest can successfully handle missing values, particularly in datasets including different types of variables. In our comparative study, missForest outperforms other methods of imputation esp. in data settings where complex interactions and non-linear relations are suspected. The out-of-bag imputation error ests. of missForest prove to be adequate in all settings. Addnl., missForest exhibits attractive computational efficiency and can cope with high-dimensional data.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC3MXhs1yms7fO&md5=44a690989dad7424d50a1662f5de2625
  33. 33
    Josse, J.; Husson, F. Handling missing values in exploratory multivariate data analysis methods. J. Soc. Fr. Stat. 2012, 153, 79– 99
    Google Scholar
    There is no corresponding record for this reference.
  34. 34
    Khoonsari, P. E.; Häggmark, A.; Lönnberg, M.; Mikus, M.; Kilander, L.; Lannfelt, L.; Bergquist, J.; Ingelsson, M.; Nilsson, P.; Kultima, K.; Shevchenko, G. Analysis of the Cerebrospinal Fluid Proteome in Alzheimer’s Disease. PLoS One 2016, 11, e0150672  DOI: 10.1371/journal.pone.0150672
    [Crossref], [PubMed], [CAS], Google Scholar
    34
    Analysis of the cerebrospinal fluid proteome in Alzheimer's disease
    Khoonsari, Payam Emami; Haggmark, Anna; Lonnberg, Maria; Mikus, Maria; Kilander, Lena; Lannfelt, Lars; Bergquist, Jonas; Ingelsson, Martin; Nilsson, Peter; Kultima, Kim; Shevchenko, Ganna
    PLoS One (2016), 11 (3), e0150672/1-e0150672/25CODEN: POLNCL; ISSN:1932-6203. (Public Library of Science)
    Alzheimer's disease is a neurodegenerative disorder accounting for more than 50% of cases of dementia. Diagnosis of Alzheimer's disease relies on cognitive tests and anal. of amyloid beta, protein tau, and hyperphosphorylated tau in cerebrospinal fluid. Although these markers provide relatively high sensitivity and specificity for early disease detection, they are not suitable for monitor of disease progression. In the present study, we used label-free shotgun mass spectrometry to analyze the cerebrospinal fluid proteome of Alzheimer's disease patients and non-demented controls to identify potential biomarkers for Alzheimer's disease. We processed the data using five programs (DecyderMS, Maxquant, OpenMS, PEAKS, and Sieve) and compared their results by means of reproducibility and peptide identification, including three different normalization methods. After depletion of high abundant proteins we found that Alzheimer's disease patients had lower fraction of low-abundance proteins in cerebrospinal fluid compared to healthy controls (p<0.05). Consequently, global normalization was found to be less accurate compared to using spiked-in chicken ovalbumin for normalization. In addn., we detd. that Sieve and OpenMS resulted in the highest reproducibility and PEAKS was the programs with the highest identification performance. Finally, we successfully verified significantly lower levels (p<0.05) of eight proteins (A2GL, APOM, C1QB, C1QC, C1S, FBLN3, PTPRZ, and SEZ6) in Alzheimer's disease compared to controls using an antibody-based detection method. These proteins are involved in different biol. roles spanning from cell adhesion and migration, to regulation of the synapse and the immune system.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC28XhtVart7vO&md5=ef30cb763212be43089453d2a5117f7e
  35. 35
    Sturm, M.; Bertsch, A.; Gröpl, C.; Hildebrandt, A.; Hussong, R.; Lange, E.; Pfeifer, N.; Schulz-Trieglaff, O.; Zerck, A.; Reinert, K.; Kohlbacher, O. OpenMS - an open-source software framework for mass spectrometry. BMC Bioinf. 2008, 9, 163  DOI: 10.1186/1471-2105-9-163
    [Crossref], [PubMed], [CAS], Google Scholar
    35
    OpenMS - an open-source software framework for mass spectrometry
    Sturm Marc; Bertsch Andreas; Gropl Clemens; Hildebrandt Andreas; Hussong Rene; Lange Eva; Pfeifer Nico; Schulz-Trieglaff Ole; Zerck Alexandra; Reinert Knut; Kohlbacher Oliver
    BMC bioinformatics (2008), 9 (), 163 ISSN:.
    BACKGROUND: Mass spectrometry is an essential analytical technique for high-throughput analysis in proteomics and metabolomics. The development of new separation techniques, precise mass analyzers and experimental protocols is a very active field of research. This leads to more complex experimental setups yielding ever increasing amounts of data. Consequently, analysis of the data is currently often the bottleneck for experimental studies. Although software tools for many data analysis tasks are available today, they are often hard to combine with each other or not flexible enough to allow for rapid prototyping of a new analysis workflow. RESULTS: We present OpenMS, a software framework for rapid application development in mass spectrometry. OpenMS has been designed to be portable, easy-to-use and robust while offering a rich functionality ranging from basic data structures to sophisticated algorithms for data analysis. This has already been demonstrated in several studies. CONCLUSION: OpenMS is available under the Lesser GNU Public License (LGPL) from the project website at http://www.openms.de.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A280%3ADC%252BD1c3ltF2ktg%253D%253D&md5=99727c3e786a099fdd18786d6e642cd8
  36. 36
    Pursiheimo, A.; Vehmas, A. P.; Afzal, S.; Suomi, T.; Chand, T.; Strauss, L.; Poutanen, M.; Rokka, A.; Corthals, G. L.; Elo, L. L. Optimization of Statistical Methods Impact on Quantitative Proteomics Data. J. Proteome Res. 2015, 14, 4118– 4126,  DOI: 10.1021/acs.jproteome.5b00183
    [ACS Full Text ACS Full Text], [CAS], Google Scholar
    36
    Optimization of Statistical Methods Impact on Quantitative Proteomics Data
    Pursiheimo, Anna; Vehmas, Anni P.; Afzal, Saira; Suomi, Tomi; Chand, Thaman; Strauss, Leena; Poutanen, Matti; Rokka, Anne; Corthals, Garry L.; Elo, Laura L.
    Journal of Proteome Research (2015), 14 (10), 4118-4126CODEN: JPROBS; ISSN:1535-3893. (American Chemical Society)
    As tools for quant. label-free mass spectrometry (MS) rapidly develop, a consensus about the best practices is not apparent. In the work described here, we compared popular statistical methods for detecting differential protein expression from quant. MS data using both controlled expts. with known quant. differences for specific proteins used as stds. as well as "real" expts. where differences in protein abundance are not known a priori. Our results suggest that data-driven reproducibility-optimization can consistently produce reliable differential expression rankings for label-free proteome tools and are straightforward in their application.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2MXhsVWitL%252FI&md5=0c2a241e97545ea49acff53fdcf8c5ec
  37. 37
    Govaert, E.; Van Steendam, K.; Scheerlinck, E.; Vossaert, L.; Meert, P.; Stella, M.; Willems, S.; De Clerck, L.; Dhaenens, M.; Deforce, D. Extracting histones for the specific purpose of label-free MS. Proteomics 2016, 16, 2937– 2944,  DOI: 10.1002/pmic.201600341
    [Crossref], [PubMed], [CAS], Google Scholar
    37
    Extracting histones for the specific purpose of label-free MS
    Govaert, Elisabeth; Van Steendam, Katleen; Scheerlinck, Ellen; Vossaert, Liesbeth; Meert, Paulien; Stella, Martina; Willems, Sander; De Clerck, Laura; Dhaenens, Maarten; Deforce, Dieter
    Proteomics (2016), 16 (23), 2937-2944CODEN: PROTC7; ISSN:1615-9853. (Wiley-VCH Verlag GmbH & Co. KGaA)
    Extg. histones from cells is the first step in studies that aim to characterize histones and their post-translational modifications (hPTMs) with MS. In the last decade, label-free quantification is more frequently being used for MS-based histone characterization. However, many histone extn. protocols were not specifically designed for label-free MS. While label-free quantification has its advantages, it is also very susceptible to tech. variation. Here, we adjust an established histone extn. protocol according to general label-free MS guidelines with a specific focus on minimizing sample handling. These protocols are first evaluated using SDS-PAGE. Hereafter, a selection of extn. protocols was used in a complete histone workflow for label-free MS. All protocols display nearly identical relative quantification of hPTMs. We thus show that, depending on the cell type under investigation and at the cost of some addnl. contaminating proteins, minimizing sample handling can be done during histone isolation. This allows analyzing bigger sample batches, leads to reduced tech. variation and minimizes the chance of in vitro alterations to the hPTM snapshot. Overall, these results allow researchers to det. the best protocol depending on the resources and goal of their specific study. Data are available via ProteomeXchange with identifier PXD002885.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC28XitVSrurzK&md5=682a99c1991f6a9ac6517da0755212a7
  38. 38
    Calf, O. W.; van Dam, N. M.; Weinhold, A.; Huber, H.; Peters, J. L. MTBLS738: Glycoalkaloid composition explains variation in slug resistance in Solanum dulcamara. Oecologia 2018, 187, 495– 506,  DOI: 10.1007/s00442-018-4064-z
    [Crossref], [PubMed], [CAS], Google Scholar
    38
    Glycoalkaloid composition explains variation in slug resistance in Solanum dulcamara
    Calf Onno W; van Dam Nicole M; Huber Heidrun; Peters Janny L; Weinhold Alexander; van Dam Nicole M; van Dam Nicole M
    Oecologia (2018), 187 (2), 495-506 ISSN:.
    In natural environments, plants have to deal with a wide range of different herbivores whose communities vary in time and space. It is believed that the chemical diversity within plant species has mainly arisen from selection pressures exerted by herbivores. So far, the effects of chemical diversity on plant resistance have mostly been assessed for arthropod herbivores. However, also gastropods, such as slugs, can cause extensive damage to plants. Here we investigate to what extent individual Solanum dulcamara plants differ in their resistance to slug herbivory and whether this variation can be explained by differences in secondary metabolites. We performed a series of preference assays using the grey field slug (Deroceras reticulatum) and S. dulcamara accessions from eight geographically distinct populations from the Netherlands. Significant and consistent variation in slug preference was found for individual accessions within and among populations. Metabolomic analyses showed that variation in steroidal glycoalkaloids (GAs) correlated with slug preference; accessions with high GA levels were consistently less damaged by slugs. One, strongly preferred, accession with particularly low GA levels contained high levels of structurally related steroidal compounds. These were conjugated with uronic acid instead of the glycoside moieties common for Solanum GAs. Our results illustrate how intraspecific variation in steroidal glycoside profiles affects resistance to slug feeding. This suggests that also slugs should be considered as important drivers in the co-evolution between plants and herbivores.
    >> More from SciFinder ®
    https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A280%3ADC%252BC1Mvls1Gguw%253D%253D&md5=172dd50c42901c7f65018fa75fab07b3

Cited By

This article is cited by 3 publications.

  1. Yannis Schumann, Julia E. Neumann, Philipp Neumann. Robust classification using average correlations as features (ACF). BMC Bioinformatics 2023, 24 (1) https://doi.org/10.1186/s12859-023-05224-0
  2. Janine Egert, Clemens Kreutz. Rcall: An R interface for MATLAB. SoftwareX 2023, 21 , 101276. https://doi.org/10.1016/j.softx.2022.101276
  3. Christophe Vanderaa, Laurent Gatto. Replication of single-cell proteomics data reveals important computational challenges. Expert Review of Proteomics 2021, 18 (10) , 835-843. https://doi.org/10.1080/14789450.2021.1988571
Download PDF

  • Abstract

    High Resolution Image
    Download MS PowerPoint Slide

    Figure 1

    Figure 1. DIMA analysis pipeline illustrated on an LC-MS/MS data set. (24) The data is sorted from top to bottom according to the frequency of MVs and the mean intensity of the proteins. Likewise, the reference data R is sorted according to the mean protein intensities after considering pattern PR of step 3. (1) The pattern PO of MVs is learned by logistic regression using the protein and sample as factorial predictors and the mean protein intensity as a continuous predictor. (2) A reference data R with few MVs is defined. (3) Various patterns PR of MVs are generated by the logistic regression model, and the respective coefficients of step 1 are incorporated into the reference data R. (4) Boxplots of the absolute imputation errors for multiple imputation algorithms. The circle indicates the median imputation deviation. The algorithms are ranked by their overall root mean square error (RMSE, red diamond). The algorithms can be divided into well-performing algorithms with an RMSE < 0.5 (green), medium performance with 0.5 < RMSE < 3 (yellow), and bad performance with RMSE > 3 (red). (5) The best-performing imputation algorithm on R (in this example impSeqRob) is recommended for the original data O and imputation of O is conducted.

    High Resolution Image
    Download MS PowerPoint Slide

    Figure 2

    Figure 2. DIMA is applied and evaluated on 142 PRIDE data sets. (A) Nine algorithms compete for being recommended as the best-performing algorithm. The R package rrcovNA with its algorithms impSeqRob (47%) and impSeq (25%) is selected most frequently, followed by missForest in 13% and imputePCA in 10%. For 5% of the Pride data sets, another algorithm is suggested. (B) The rank of the imputation algorithms obtained in the 142 PRIDE data sets is shown as a box plot. The seven algorithms with the lowest median rank are also the seven most frequently selected algorithms by DIMA (A). The algorithms with a median rank lower than 5% are highlighted in green, and algorithms with a median rank greater than 20 are highlighted in red.

    High Resolution Image
    Download MS PowerPoint Slide

    Figure 3

    Figure 3. Performance of DIMA is evaluated on simulated data S with the incorporation of various proportions of MV and MNAR/MCAR ratios. The RMSE (color-coded) and rank (first entry) obtained by the best-performing imputation algorithm recommended by DIMA (second entry) compared to direct imputation assessment (third entry) over 500 data simulations are calculated. The algorithm recommended by DIMA is within the top three out of 27 approaches in all cases. For MV < 20% (A), the additive regression aregImpute with type regression (reg) outperforms, between 20 and 30% MVs; (B) several algorithms compete against each other and for MV > 30%; (C) the random forest algorithm missForest performs best.

    High Resolution Image
    Download MS PowerPoint Slide

  • References

    ARTICLE SECTIONS
    Jump To

    This article references 38 other publications.

    1. 1
      McGurk, K. A. The use of missing values in proteomic data-independent acquisition mass spectrometry to enable disease activity discrimination. Bioinformatics 2020, 36, 2217– 2223,  DOI: 10.1093/bioinformatics/btz898
      [Crossref], [PubMed], [CAS], Google Scholar
      1
      The use of missing values in proteomic data-independent acquisition mass spectrometry to enable disease activity discrimination
      McGurk, Kathryn A.; Dagliati, Arianna; Chiasserini, Davide; Lee, Dave; Plant, Darren; Baricevic-Jones, Ivona; Kelsall, Janet; Eineman, Rachael; Reed, Rachel; Geary, Bethany; Unwin, Richard D.; Nicolaou, Anna; Keavney, Bernard D.; Barton, Anne; Whetton, Anthony D.; Geifman, Nophar
      Bioinformatics (2020), 36 (7), 2217-2223CODEN: BOINFP; ISSN:1367-4811. (Oxford University Press)
      Data-independent acquisition mass spectrometry allows for comprehensive peptide detection and relative quantification than std. data-dependent approaches. While less prone to missing values, these still exist. Current approaches for handling the so-called missingness have challenges. We hypothesized that non-random missingness is a useful biol. measure and demonstrate the importance of analyzing missingness for proteomic discovery within a longitudinal study of disease activity. The magnitude of missingness did not correlate with mean peptide concn. The magnitude of missingness for each protein strongly correlated between collection time points (baseline, 3 mo, 6 mo; R = 0.95-0.97, confidence interval = 0.94-0.97) indicating little time-dependent effect. This allowed for the identification of proteins with outlier levels of missingness that differentiate between the patient groups characterized by different patterns of disease activity. The assocn. of these proteins with disease activity was confirmed by machine learning techniques. Our novel approach complements analyses on complete observations and other missing value strategies in biomarker prediction of disease activity.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3cXitlOmu7bI&md5=d2a3cd6644ada2086f6aaa6722145f68
    2. 2
      Poulos, R. C. Strategies to enable large-scale proteomics for reproducible research. Nat. Commun. 2020, 11, 3793  DOI: 10.1038/s41467-020-17641-3
      [Crossref], [PubMed], [CAS], Google Scholar
      2
      Strategies to enable large-scale proteomics for reproducible research
      Poulos, Rebecca C.; Hains, Peter G.; Shah, Rohan; Lucas, Natasha; Xavier, Dylan; Manda, Srikanth S.; Anees, Asim; Koh, Jennifer M. S.; Mahboob, Sadia; Wittman, Max; Williams, Steven G.; Sykes, Erin K.; Hecker, Michael; Dausmann, Michael; Wouters, Merridee A.; Ashman, Keith; Yang, Jean; Wild, Peter J.; deFazio, Anna; Balleine, Rosemary L.; Tully, Brett; Aebersold, Ruedi; Speed, Terence P.; Liu, Yansheng; Reddel, Roger R.; Robinson, Phillip J.; Zhong, Qing
      Nature Communications (2020), 11 (1), 3793CODEN: NCAOBW; ISSN:2041-1723. (Nature Research)
      Abstr.: Reproducible research is the bedrock of exptl. science. To enable the deployment of large-scale proteomics, we assess the reproducibility of mass spectrometry (MS) over time and across instruments and develop computational methods for improving quant. accuracy. We perform 1560 data independent acquisition (DIA)-MS runs of eight samples contg. known proportions of ovarian and prostate cancer tissue and yeast, or control HEK293T cells. Replicates are run on six mass spectrometers operating continuously with varying maintenance schedules over four months, interspersed with ∼5000 other runs. We utilize neg. controls and replicates to remove unwanted variation and enhance biol. signal, outperforming existing methods. We also design a method for reducing missing values. Integrating these computational modules into a pipeline (ProNorM), we mitigate variation among instruments over time and accurately predict tissue proportions. We demonstrate how to improve the quant. anal. of large-scale DIA-MS data, providing a pathway toward clin. proteomics.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3cXhsFajsb7M&md5=848bd6df41e1fc16efef3d87f76de510
    3. 3
      Brenes, A.; Hukelmann, J.; Bensaddek, D.; Lamond, A. I. Multibatch TMT Reveals False Positives, Batch Effects and Missing Values. Mol. Cell. Proteomics 2019, 18, 1967– 1980,  DOI: 10.1074/mcp.RA119.001472
      [Crossref], [PubMed], [CAS], Google Scholar
      3
      Multibatch TMT reveals false positives, batch effects and missing values
      Brenes, Alejandro; Hukelmann, Jens; Bensaddek, Dalila; Lamond, Angus I.
      Molecular & Cellular Proteomics (2019), 18 (10), 1967-1980CODEN: MCPOBS; ISSN:1535-9484. (American Society for Biochemistry and Molecular Biology)
      Multiplexing strategies for large-scale proteomic analyses have become increasingly prevalent, tandem mass tags (TMT) in particular. Here we used a large iPSC proteomic expt. with twenty-four 10-plex TMT batches to evaluate the effect of integrating multiple TMT batches within a single anal. We identified a significant inflation rate of protein missing values as multiple batches are integrated and show that this pattern is aggravated at the peptide level. We also show that without normalization strategies to address the batch effects, the high precision of quantitation within a single multiplexed TMT batch is not reproduced when data from multiple TMT batches are integrated. Further, the incidence of false positives was studied by using Y chromosome peptides as an internal control. The iPSC lines quantified in this data set were derived from both male and female donors, hence the peptides mapped to the Y chromosome should be absent from female lines. Nonetheless, these Y chromosome-specific peptides were consistently detected in the female channels of all TMT batches. We then used the same Y chromosome specific peptides to quantify the level of ion coisolation as well as the effect of primary and secondary reporter ion interference. These results were used to propose solns. to mitigate the limitations of multi-batch TMT analyses. We confirm that including a common ref. line in every batch increases precision by facilitating normalization across the batches and we propose exptl. designs that minimize the effect of cross population reporter ion interference.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC1MXitlSjtb3J&md5=1321490bdeb894c40aa34eff5d12546b
    4. 4
      Wei, R.; Wang, J.; Su, M.; Jia, E.; Chen, S.; Chen, T.; Ni, Y. Missing Value Imputation Approach for Mass Spectrometry-based Metabolomics Data. Sci. Rep. 2018, 8, 663  DOI: 10.1038/s41598-017-19120-0
      [Crossref], [PubMed], [CAS], Google Scholar
      4
      Missing Value Imputation Approach for Mass Spectrometry-based Metabolomics Data
      Wei Runmin; Wang Jingye; Su Mingming; Chen Shaoqiu; Ni Yan; Wei Runmin; Chen Shaoqiu; Su Mingming; Jia Erik; Chen Tianlu
      Scientific reports (2018), 8 (1), 663 ISSN:.
      Missing values exist widely in mass-spectrometry (MS) based metabolomics data. Various methods have been applied for handling missing values, but the selection can significantly affect following data analyses. Typically, there are three types of missing values, missing not at random (MNAR), missing at random (MAR), and missing completely at random (MCAR). Our study comprehensively compared eight imputation methods (zero, half minimum (HM), mean, median, random forest (RF), singular value decomposition (SVD), k-nearest neighbors (kNN), and quantile regression imputation of left-censored data (QRILC)) for different types of missing values using four metabolomics datasets. Normalized root mean squared error (NRMSE) and NRMSE-based sum of ranks (SOR) were applied to evaluate imputation accuracy. Principal component analysis (PCA)/partial least squares (PLS)-Procrustes analysis were used to evaluate the overall sample distribution. Student's t-test followed by correlation analysis was conducted to evaluate the effects on univariate statistics. Our findings demonstrated that RF performed the best for MCAR/MAR and QRILC was the favored one for left-censored MNAR. Finally, we proposed a comprehensive strategy and developed a public-accessible web-tool for the application of missing value imputation in metabolomics ( https://metabolomics.cc.hawaii.edu/software/MetImp/ ).
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A280%3ADC%252BC1MvgtlyrsQ%253D%253D&md5=f161109ca8155e59a728dd2761e994f6
    5. 5
      Webb-Robertson, B.-J. M.; Wiberg, H. K.; Matzke, M. M.; Brown, J. N.; Wang, J.; McDermott, J. E.; Smith, R. D.; Rodland, K. D.; Metz, T. O.; Pounds, J. G.; Waters, K. M. Reviewand Evaluationand and Discussion of the Challenges of Missing Value Imputation for Mass Spectrometry-Based Label-Free Global Proteomics. J. Proteome Res. 2015, 14, 1993– 2001,  DOI: 10.1021/pr501138h
      [ACS Full Text ACS Full Text], [CAS], Google Scholar
      5
      Review, Evaluation, and Discussion of the Challenges of Missing Value Imputation for Mass Spectrometry-Based Label-Free Global Proteomics
      Webb-Robertson, Bobbie-Jo M.; Wiberg, Holli K.; Matzke, Melissa M.; Brown, Joseph N.; Wang, Jing; McDermott, Jason E.; Smith, Richard D.; Rodland, Karin D.; Metz, Thomas O.; Pounds, Joel G.; Waters, Katrina M.
      Journal of Proteome Research (2015), 14 (5), 1993-2001CODEN: JPROBS; ISSN:1535-3893. (American Chemical Society)
      A review. In this review, we apply selected imputation strategies to label-free liq. chromatog.-mass spectrometry (LC-MS) proteomics datasets to evaluate the accuracy with respect to metrics of variance and classification. We evaluate several commonly used imputation approaches for individual merits and discuss the caveats of each approach with respect to the example LC-MS proteomics data. In general, local similarity-based approaches, such as the regularized expectation maximization and least-squares adaptive algorithms, yield the best overall performances with respect to metrics of accuracy and robustness. However, no single algorithm consistently outperforms the remaining approaches, and in some cases, performing classification without imputation sometimes yielded the most accurate classification. Thus, because of the complex mechanisms of missing data in proteomics, which also vary from peptide to protein, no individual method is a single soln. for imputation. On the basis of the observations in this review, the goal for imputation in the field of computational proteomics should be to develop new approaches that work generically for this data type and new strategies to guide users in the selection of the best imputation for their dataset and anal. objectives.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2MXmtV2ltL0%253D&md5=218ce277ccefd57aab2ad11987b943c6
    6. 6
      Lazar, C.; Laurent, G.; Myriam, F.; Christophe, B.; Thomas, B. Accounting for the Multiple Natures of Missing Values in Label-Free Quantitative Proteomics Data Sets to Compare Imputation Strategies. J. Proteome Res. 2016, 15, 1116– 1125,  DOI: 10.1021/acs.jproteome.5b00981
      [ACS Full Text ACS Full Text], [CAS], Google Scholar
      6
      Accounting for the Multiple Natures of Missing Values in Label-Free Quantitative Proteomics Data Sets to Compare Imputation Strategies
      Lazar, Cosmin; Gatto, Laurent; Ferro, Myriam; Bruley, Christophe; Burger, Thomas
      Journal of Proteome Research (2016), 15 (4), 1116-1125CODEN: JPROBS; ISSN:1535-3893. (American Chemical Society)
      Missing values are a genuine issue in label-free quant. proteomics. Recent works have surveyed the different statistical methods to conduct imputation and have compared them on real or simulated data sets and recommended a list of missing value imputation methods for proteomics application. Although insightful, these comparisons do not account for two important facts: (i) depending on the proteomics data set, the missingness mechanism may be of different natures and (ii) each imputation method is devoted to a specific type of missingness mechanism. As a result, we believe that the question at stake is not to find the most accurate imputation method in general but instead the most appropriate one. We describe a series of comparisons that support our views: For instance, we show that a supposedly "under-performing" method (i.e., giving baseline av. results), if applied at the "appropriate" time in the data-processing pipeline (before or after peptide aggregation) on a data set with the "appropriate" nature of missing values, can outperform a blindly applied, supposedly "better-performing" method (i.e., the ref. method from the state-of-the-art). This leads us to formulate few practical guidelines regarding the choice and the application of an imputation method in a proteomics context.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC28XivFOntbw%253D&md5=edecca833d3f9183fb1fe3e9ff14e8af
    7. 7
      Rubin, D. B. Inference and missing data. Biometrika 1976, 63, 581– 592,  DOI: 10.1093/biomet/63.3.581
      [Crossref], Google Scholar
      There is no corresponding record for this reference.
    8. 8
      Karpievitch, Y. V.; Dabney, A. R.; Smith, R. D. Normalization and missing value imputation for label-free LC-MS analysis. BMC Bioinf. 2012, 13, S5  DOI: 10.1186/1471-2105-13-S16-S5
      [Crossref], [PubMed], [CAS], Google Scholar
      8
      Normalization and missing value imputation for label-free LC-MS analysis
      Karpievitch, Yuliya V.; Dabney, Alan R.; Smith, Richard D.
      BMC Bioinformatics (2012), 13 (Suppl. 16), S5CODEN: BBMIC4; ISSN:1471-2105. (BioMed Central Ltd.)
      Shotgun proteomic data are affected by a variety of known and unknown systematic biases as well as high proportions of missing values. Typically, normalization is performed in an attempt to remove systematic biases from the data before statistical inference, sometimes followed by missing value imputation to obtain a complete matrix of intensities. Here we discuss several approaches to normalization and dealing with missing values, some initially developed for microarray data and some developed specifically for mass spectrometry-based data.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC38XhvVyis7bN&md5=f1dbc6f041dba77202a072e3e0f4a959
    9. 9
      Välikangas, T.; Suomi, T.; Elo, L. L. A comprehensive evaluation of popular proteomics software workflows for label-free proteome quantification and imputation. Briefings Bioinf. 2017, 19, 1344– 1355,  DOI: 10.1093/bib/bbx054
      [Crossref], Google Scholar
      There is no corresponding record for this reference.
    10. 10
      Wang, J.; Li, L.; Chen, T.; Ma, J.; Zhu, Y.; Zhuang, J.; Chang, C. In-depth method assessments of differentially expressed protein detection for shotgun proteomics data with missing values. Sci. Rep. 2017, 7, 3367  DOI: 10.1038/s41598-017-03650-8
      [Crossref], [PubMed], [CAS], Google Scholar
      10
      In-depth method assessments of differentially expressed protein detection for shotgun proteomics data with missing values
      Wang Jinxia; Li Liwei; Chen Tao; Ma Jie; Zhu Yunping; Chang Cheng; Wang Jinxia; Zhuang Jujuan; Wang Jinxia
      Scientific reports (2017), 7 (1), 3367 ISSN:.
      Considering as one of the major goals in quantitative proteomics, detection of the differentially expressed proteins (DEPs) plays an important role in biomarker selection and clinical diagnostics. There have been plenty of algorithms and tools focusing on DEP detection in proteomics research. However, due to the different application scopes of these methods, and various kinds of experiment designs, it is not very apparent about the best choice for large-scale proteomics data analyses. Moreover, given the fact that proteomics data usually contain high percentage of missing values (MVs), but few replicates, a systematic evaluation of the DEP detection methods combined with the MV imputation methods is essential and urgent. Here, we analyzed a total of four representative imputation methods and five DEP methods on different experimental and simulated datasets. The results showed that (i) MV imputation could not always improve the performances of DEP detection methods and the imputation effects differed in the missing value percentages; (ii) the DEP detection methods had different statistical powers affected by the percentage of MVs. Two statistical methods (i.e. the empirical Bayesian random censoring threshold model, and the significance analysis of microarray) performed better than the other evaluated methods in terms of accuracy and sensitivity.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A280%3ADC%252BC1cnnvF2iug%253D%253D&md5=038103fc16b08f4e3b802f37c1172b39
    11. 11
      Janssen, K. J.; Donders, A. R. T.; Harrell, F. E.; Vergouwe, Y.; Chen, Q.; Grobbee, D. E.; Moons, K. G. Missing covariate data in medical research: To impute is better than to ignore. J. Clin. Epidemiol. 2010, 63, 721– 727,  DOI: 10.1016/j.jclinepi.2009.12.008
      [Crossref], [PubMed], [CAS], Google Scholar
      11
      Missing covariate data in medical research: to impute is better than to ignore
      Janssen Kristel J M; Donders A Rogier T; Harrell Frank E Jr; Vergouwe Yvonne; Chen Qingxia; Grobbee Diederick E; Moons Karel G M
      Journal of clinical epidemiology (2010), 63 (7), 721-7 ISSN:.
      OBJECTIVE: We compared popular methods to handle missing data with multiple imputation (a more sophisticated method that preserves data). STUDY DESIGN AND SETTING: We used data of 804 patients with a suspicion of deep venous thrombosis (DVT). We studied three covariates to predict the presence of DVT: d-dimer level, difference in calf circumference, and history of leg trauma. We introduced missing values (missing at random) ranging from 10% to 90%. The risk of DVT was modeled with logistic regression for the three methods, that is, complete case analysis, exclusion of d-dimer level from the model, and multiple imputation. RESULTS: Multiple imputation showed less bias in the regression coefficients of the three variables and more accurate coverage of the corresponding 90% confidence intervals than complete case analysis and dropping d-dimer level from the analysis. Multiple imputation showed unbiased estimates of the area under the receiver operating characteristic curve (0.88) compared with complete case analysis (0.77) and when the variable with missing values was dropped (0.65). CONCLUSION: As this study shows that simple methods to deal with missing data can lead to seriously misleading results, we advise to consider multiple imputation. The purpose of multiple imputation is not to create data, but to prevent the exclusion of observed data.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A280%3ADC%252BC3czmtV2gtQ%253D%253D&md5=bab541926c308b8b5c901d42982dae27
    12. 12
      Brock, G. N.; Shaffer, J. R.; Blakesley, R. E.; Lotz, M. J.; Tseng, G. C. Which missing value imputation method to use in expression profiles: a comparative study and two selection schemes. BMC Bioinf. 2008, 9, 12  DOI: 10.1186/1471-2105-9-12
      [Crossref], [PubMed], [CAS], Google Scholar
      12
      Which missing value imputation method to use in expression profiles: a comparative study and two selection schemes
      Brock Guy N; Shaffer John R; Blakesley Richard E; Lotz Meredith J; Tseng George C
      BMC bioinformatics (2008), 9 (), 12 ISSN:.
      BACKGROUND: Gene expression data frequently contain missing values, however, most down-stream analyses for microarray experiments require complete data. In the literature many methods have been proposed to estimate missing values via information of the correlation patterns within the gene expression matrix. Each method has its own advantages, but the specific conditions for which each method is preferred remains largely unclear. In this report we describe an extensive evaluation of eight current imputation methods on multiple types of microarray experiments, including time series, multiple exposures, and multiple exposures x time series data. We then introduce two complementary selection schemes for determining the most appropriate imputation method for any given data set. RESULTS: We found that the optimal imputation algorithms (LSA, LLS, and BPCA) are all highly competitive with each other, and that no method is uniformly superior in all the data sets we examined. The success of each method can also depend on the underlying "complexity" of the expression data, where we take complexity to indicate the difficulty in mapping the gene expression matrix to a lower-dimensional subspace. We developed an entropy measure to quantify the complexity of expression matrixes and found that, by incorporating this information, the entropy-based selection (EBS) scheme is useful for selecting an appropriate imputation algorithm. We further propose a simulation-based self-training selection (STS) scheme. This technique has been used previously for microarray data imputation, but for different purposes. The scheme selects the optimal or near-optimal method with high accuracy but at an increased computational cost. CONCLUSION: Our findings provide insight into the problem of which imputation method is optimal for a given data set. Three top-performing methods (LSA, LLS and BPCA) are competitive with each other. Global-based imputation methods (PLS, SVD, BPCA) performed better on mcroarray data with lower complexity, while neighbour-based methods (KNN, OLS, LSA, LLS) performed better in data with higher complexity. We also found that the EBS and STS schemes serve as complementary and effective tools for selecting the optimal imputation algorithm.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A280%3ADC%252BD1c7jsVyhug%253D%253D&md5=550ac2936960ddcd03d9ca2586c4281d
    13. 13
      To, K. T.; Fry, R. C.; Reif, D. M. Characterizing the effects of missing data and evaluating imputation methods for chemical prioritization applications using ToxPi. BioData Min. 2018, 11, 10  DOI: 10.1186/s13040-018-0169-5
      [Crossref], [PubMed], [CAS], Google Scholar
      13
      Characterizing the effects of missing data and evaluating imputation methods for chemical prioritization applications using ToxPi
      To Kimberly T; Reif David M; Fry Rebecca C; Reif David M; Reif David M
      BioData mining (2018), 11 (), 10 ISSN:1756-0381.
      BACKGROUND: The Toxicological Priority Index (ToxPi) is a method for prioritization and profiling of chemicals that integrates data from diverse sources. However, individual data sources ("assays"), such as in vitro bioassays or in vivo study endpoints, often feature sections of missing data, wherein subsets of chemicals have not been tested in all assays. In order to investigate the effects of missing data and recommend solutions, we designed simulation studies around high-throughput screening data generated by the ToxCast and Tox21 programs on chemicals highlighted by the Agency for Toxic Substances and Disease Registry's (ATSDR) Substance Priority List (SPL), which helps prioritize environmental research and remediation resources. RESULTS: Our simulations explored a wide range of scenarios concerning data (0-80% assay data missing per chemical), modeling (ToxPi models containing from 160-700 different assays), and imputation method (k-Nearest-Neighbor, Max, Mean, Min, Binomial, Local Least Squares, and Singular Value Decomposition). We find that most imputation methods result in significant changes to ToxPi score, except for datasets with a small number of assays. If we consider rank change conditional on these significant changes to ToxPi score, we find that ranks of chemicals in the minimum value imputation, SVD imputation, and kNN imputation sets are more sensitive to the score changes. CONCLUSIONS: We found that the choice of imputation strategy exerted significant influence over both scores and associated ranks, and the most sensitive scenarios were those involving fewer assays plus higher proportions of missing data. By characterizing the effects of missing data and the relative benefit of imputation approaches across real-world data scenarios, we can augment confidence in the robustness of decisions regarding the health and ecological effects of environmental chemicals.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A280%3ADC%252BB3c%252FhslOksQ%253D%253D&md5=5ea3228f441ab27190b2ada91f5059e6
    14. 14
      Poyatos, R.; Sus, O.; Badiella, L.; Mencuccini, M.; Martinez-Vilalta, J. Gap-filling a spatially explicit plant trait database: comparing imputation methods and different levels of environmental information. Biogeosciences 2018, 15, 2601– 2617,  DOI: 10.5194/bg-15-2601-2018
      [Crossref], Google Scholar
      There is no corresponding record for this reference.
    15. 15
      Lenz, M.; Schulz, A.; Koeck, T.; Rapp, S.; Nagler, M.; Sauer, M.; Eggebrecht, L.; Cate, V. T.; Panova-Noeva, M.; Prochaska, J. H.; Lackner, K. J.; Münzel, T.; Leineweber, K.; Wild, P. S.; Andrade-Navarro, M. A. Missing value imputation in proximity extension assay-based targeted proteomics data. PLoS One 2020, 15, e0243487  DOI: 10.1371/journal.pone.0243487
      [Crossref], [PubMed], [CAS], Google Scholar
      15
      Missing value imputation in proximity extension assay-based targeted proteomics data
      Lenz, Michael; Schulz, Andreas; Koeck, Thomas; Rapp, Steffen; Nagler, Markus; Sauer, Madeleine; Eggebrecht, Lisa; Ten Cate, Vincent; Panova-Noeva, Marina; Prochaska, Juergen H.; Lackner, Karl J.; Muenzel, Thomas; Leineweber, Kirsten; Wild, Philipp S.; Andrade-Navarro, Miguel A.
      PLoS One (2020), 15 (12), e0243487CODEN: POLNCL; ISSN:1932-6203. (Public Library of Science)
      Targeted proteomics utilizing antibody-based proximity extension assays provides sensitive and highly specific quantifications of plasma protein levels. Multivariate anal. of this data is hampered by frequent missing values (random or left censored), calling for imputation approaches. While appropriate missing-value imputation methods exist, benchmarks of their performance in targeted proteomics data are lacking. Here, we assessed the performance of two methods for imputation of values missing completely at random, the previously top-benchmarked 'missForest' and the recently published 'GSimp' method. Evaluation was accomplished by comparing imputed with remeasured relative concns. of 91 inflammation related circulating proteins in 86 samples from a cohort of 645 patients with venous thromboembolism. The median Pearson correlation between imputed and remeasured protein expression values was 69.0% for missForest and 71.6% for GSimp (p = 5.8e-4). Imputation with missForest resulted in stronger redn. of variance compared to GSimp (median relative variance of 25.3% vs. 68.6%, p = 2.4e-16) and undesired larger bias in downstream anal. Irresp. of the imputation method used, the 91 imputed proteins revealed large variations in imputation accuracy, driven by differences in signal to noise ratio and information overlap between proteins. In summary, GSimp outperformed missForest, while both methods show good overall imputation accuracy with large variations between proteins.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3cXis1GgsrrI&md5=a13ae3b2e3fc861f525e13f71405736d
    16. 16
      Bramer, L. M.; Irvahn, J.; Piehowski, P. D.; Rodland, K. D.; Webb-Robertson, B.-J. M. A Review of Imputation Strategies for Isobaric Labeling-Based Shotgun Proteomics. J. Proteome Res. 2021, 20, 1– 13,  DOI: 10.1021/acs.jproteome.0c00123
      [ACS Full Text ACS Full Text], [CAS], Google Scholar
      16
      A Review of Imputation Strategies for Isobaric Labeling-Based Shotgun Proteomics
      Bramer, Lisa M.; Irvahn, Jan; Piehowski, Paul D.; Rodland, Karin D.; Webb-Robertson, Bobbie-Jo M.
      Journal of Proteome Research (2021), 20 (1), 1-13CODEN: JPROBS; ISSN:1535-3893. (American Chemical Society)
      A review. The throughput efficiency and increased depth of coverage provided by isobaric-labeled proteomics measurements have led to increased usage of these techniques. However, the structure of missing data is different than unlabeled studies, which prompts the need for this review to compare the efficacy of nine imputation methods on large isobaric-labeled proteomics data sets to guide researchers on the appropriateness of various imputation methods. Imputation methods were evaluated by accuracy, statistical hypothesis test inference, and run time. In general, expectation maximization and random forest imputation methods yielded the best performance, and const.-based methods consistently performed poorly across all data set sizes and percentages of missing values. For data sets with small sample sizes and higher percentages of missing data, results indicate that statistical inference with no imputation may be preferable. On the basis of the findings in this review, there are core imputation methods that perform better for isobaric-labeled proteomics data, but great care and consideration as to whether imputation is the optimal strategy should be given for data sets comprised of a small no. of samples.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3cXhvVerur3N&md5=6de86677050965321a2b7a05314cb1f7
    17. 17
      de Souto, M. C. P.; Jaskowiak, P. A.; Costa, I. G. Impact of missing data imputation methods on gene expression clustering and classification. BMC Bioinf. 2015, 16, 64  DOI: 10.1186/s12859-015-0494-3
      [Crossref], [PubMed], [CAS], Google Scholar
      17
      Impact of missing data imputation methods on gene expression clustering and classification
      de Souto Marcilio C P; Jaskowiak Pablo A; Costa Ivan G; Costa Ivan G
      BMC bioinformatics (2015), 16 (), 64 ISSN:.
      BACKGROUND: Several missing value imputation methods for gene expression data have been proposed in the literature. In the past few years, researchers have been putting a great deal of effort into presenting systematic evaluations of the different imputation algorithms. Initially, most algorithms were assessed with an emphasis on the accuracy of the imputation, using metrics such as the root mean squared error. However, it has become clear that the success of the estimation of the expression value should be evaluated in more practical terms as well. One can consider, for example, the ability of the method to preserve the significant genes in the dataset, or its discriminative/predictive power for classification/clustering purposes. RESULTS AND CONCLUSIONS: We performed a broad analysis of the impact of five well-known missing value imputation methods on three clustering and four classification methods, in the context of 12 cancer gene expression datasets. We employed a statistical framework, for the first time in this field, to assess whether different imputation methods improve the performance of the clustering/classification methods. Our results suggest that the imputation methods evaluated have a minor impact on the classification and downstream clustering analyses. Simple methods such as replacing the missing values by mean or the median values performed as well as more complex strategies. The datasets analyzed in this study are available at http://costalab.org/Imputation/ .
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A280%3ADC%252BC2Mjlslemsg%253D%253D&md5=210bc346aff2d28f775b2c1f2a4093b7
    18. 18
      Liu, M.; Dongre, A. Proper imputation of missing values in proteomics datasets for differential expression analysis. Briefings Bioinf. 2020, 0, bbaa112  DOI: 10.1093/bib/bbaa112
      [Crossref], Google Scholar
      There is no corresponding record for this reference.
    19. 19
      Rodwell, L.; Lee, K. J.; Romaniuk, H.; Carlin, J. B. Comparison of methods for imputing limited-range variables: a simulation study. BMC Med. Res. Methodol. 2014, 14, 57  DOI: 10.1186/1471-2288-14-57
      [Crossref], [PubMed], [CAS], Google Scholar
      19
      Comparison of methods for imputing limited-range variables: a simulation study
      Rodwell Laura; Lee Katherine J; Romaniuk Helena; Carlin John B
      BMC medical research methodology (2014), 14 (), 57 ISSN:.
      BACKGROUND: Multiple imputation (MI) was developed as a method to enable valid inferences to be obtained in the presence of missing data rather than to re-create the missing values. Within the applied setting, it remains unclear how important it is that imputed values should be plausible for individual observations. One variable type for which MI may lead to implausible values is a limited-range variable, where imputed values may fall outside the observable range. The aim of this work was to compare methods for imputing limited-range variables, with a focus on those that restrict the range of the imputed values. METHODS: Using data from a study of adolescent health, we consider three variables based on responses to the General Health Questionnaire (GHQ), a tool for detecting minor psychiatric illness. These variables, based on different scoring methods for the GHQ, resulted in three continuous distributions with mild, moderate and severe positive skewness. In an otherwise complete dataset, we set 33% of the GHQ observations to missing completely at random or missing at random; repeating this process to create 1000 datasets with incomplete data for each scenario.For each dataset, we imputed values on the raw scale and following a zero-skewness log transformation using: univariate regression with no rounding; post-imputation rounding; truncated normal regression; and predictive mean matching. We estimated the marginal mean of the GHQ and the association between the GHQ and a fully observed binary outcome, comparing the results with complete data statistics. RESULTS: Imputation with no rounding performed well when applied to data on the raw scale. Post-imputation rounding and imputation using truncated normal regression produced higher marginal means than the complete data estimate when data had a moderate or severe skew, and this was associated with under-coverage of the complete data estimate. Predictive mean matching also produced under-coverage of the complete data estimate. For the estimate of association, all methods produced similar estimates to the complete data. CONCLUSIONS: For data with a limited range, multiple imputation using techniques that restrict the range of imputed values can result in biased estimates for the marginal mean when data are highly skewed.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A280%3ADC%252BC2cnmvVKlsA%253D%253D&md5=67b590a65a7366914e38279d9ed1cb6a
    20. 20
      Kruttika, D.; Simion, K.; R, J. M.; J, P. S. A Simple Optimization Workflow to Enable Precise and Accurate Imputation of Missing Values in Proteomic Datasets. bioRxiv 2020, 1– 39,  DOI: 10.1101/2019.12.11.123456
      [Crossref], Google Scholar
      There is no corresponding record for this reference.
    21. 21
      Jin, L.; Bi, Y.; Hu, C.; Qu, J.; Shen, S.; Wang, X.; Tian, Y. A comparative study of evaluating missing value imputation methodsin label-free proteomics. Sci. Rep. 2021, 11, 1760  DOI: 10.1038/s41598-021-81279-4
      [Crossref], [PubMed], [CAS], Google Scholar
      21
      A comparative study of evaluating missing value imputation methods in label-free proteomics
      Jin, Liang; Bi, Yingtao; Hu, Chenqi; Qu, Jun; Shen, Shichen; Wang, Xue; Tian, Yu
      Scientific Reports (2021), 11 (1), 1760CODEN: SRCEC3; ISSN:2045-2322. (Nature Research)
      The presence of missing values (MVs) in label-free quant. proteomics greatly reduces the completeness of data. Imputation has been widely utilized to handle MVs, and selection of the proper method is crit. for the accuracy and reliability of imputation. Here we present a comparative study that evaluates the performance of seven popular imputation methods with a large-scale benchmark dataset and an immune cell dataset. Simulated MVs were incorporated into the complete part of each dataset with different combinations of MV rates and missing not at random (MNAR) rates. Normalized root mean square error (NRMSE) was applied to evaluate the accuracy of protein abundances and intergroup protein ratios after imputation. Detection of true positives (TPs) and false altered-protein discovery rate (FADR) between groups were also compared using the benchmark dataset. Furthermore, the accuracy of handling real MVs was assessed by comparing enriched pathways and signature genes of cell activation after imputing the immune cell dataset. We obsd. that the accuracy of imputation is primarily affected by the MNAR rate rather than the MV rate, and downstream anal. can be largely impacted by the selection of imputation methods. A random forest-based imputation method consistently outperformed other popular methods by achieving the lowest NRMSE, high amt. of TPs with the av. FADR < 5%, and the best detection of relevant pathways and signature genes, highlighting it as the most suitable method for label-free proteomics.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3MXhvFaqtrk%253D&md5=34e7040ccb9fb0dfe9b7e371dac1bcb1
    22. 22
      Audoux, J.; Salson, M.; Grosset, C. F.; Beaumeunier, S.; Holder, J.-M.; Commes, T.; Philippe, N. SimBA: A methodology and tools for evaluating the performance of RNA-Seq bioinformatic pipelines. BMC Bioinf. 2017, 18, 428  DOI: 10.1186/s12859-017-1831-5
      [Crossref], [PubMed], [CAS], Google Scholar
      22
      SimBA: a methodology and tools for evaluating the performance of RNA-Seq bioinformatic pipelines
      Audoux, Jerome; Salson, Mikael; Grosset, Christophe F.; Beaumeunier, Sacha; Holder, Jean-Marc; Commes, Therese; Philippe, Nicolas
      BMC Bioinformatics (2017), 18 (), 428/1-428/14CODEN: BBMIC4; ISSN:1471-2105. (BioMed Central Ltd.)
      The evolution of next-generation sequencing (NGS) technologies has led to increased focus on RNA-Seq. Many bioinformatic tools have been developed for RNA-Seq anal., each with unique performance characteristics and configuration parameters. Users face an increasingly complex task in understanding which bioinformatic tools are best for their specific needs and how they should be configured. In order to provide some answers to these questions, we investigate the performance of leading bioinformatic tools designed for RNA-Seq anal. and propose a methodol. for systematic evaluation and comparison of performance to help users make well informed choices. To evaluate RNA-Seq pipelines, we developed a suite of two benchmarking tools. SimCT generates simulated datasets that get as close as possible to specific real biol. conditions accompanied by the list of genomic incidents and mutations that have been inserted. BenchCT then compares the output of any bioinformatics pipeline that has been run against a SimCT dataset with the simulated genomic and transcriptional variations it contains to give an accurate performance evaluation in addressing specific biol. question. We used these tools to simulate a real-world genomic medicine questions involving the comparison of healthy and cancerous cells. Results revealed that performance in addressing a particular biol. context varied significantly depending on the choice of tools and settings used. We also found that by combining the output of certain pipelines, substantial performance improvements could be achieved. Our research emphasizes the importance of selecting and configuring bioinformatic tools for the specific biol. question being investigated to obtain optimal results. Pipeline designers, developers and users should include benchmarking in the context of their biol. question as part of their design and quality control process. Our SimBA suite of benchmarking tools provides a reliable basis for comparing the performance of RNA-Seq bioinformatics pipelines in addressing a specific biol. question. We would like to see the creation of a ref. corpus of data-sets that would allow accurate comparison between benchmarks performed by different groups and the publication of more benchmarks based on this public corpus.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC1cXitVOhtbrM&md5=024e4f119feb752a3458f3a643e71527
    23. 23
      Wang, S.; Li, W.; Hu, L.; Cheng, J.; Yang, H.; Liu, Y. NAguideR: performing and prioritizing missing value imputations for consistent bottom-up proteomic analyses. Nucleic Acids Res. 2020, 48, e83,  DOI: 10.1093/nar/gkaa498
      [Crossref], [PubMed], [CAS], Google Scholar
      23
      NAguideR: performing and prioritizing missing value imputations for consistent bottom-up proteomic analyses
      Wang, Shisheng; Li, Wenxue; Hu, Liqiang; Cheng, Jingqiu; Yang, Hao; Liu, Yansheng
      Nucleic Acids Research (2020), 48 (14), e83CODEN: NARHAD; ISSN:1362-4962. (Oxford University Press)
      Mass spectrometry (MS)-based quant. proteomics expts. frequently generate data with missing values, which may profoundly affect downstream analyses. A wide variety of imputation methods have been established to deal with the missing-value issue. To date, however, there is a scarcity of efficient, systematic, and easy-to-handle tools that are tailored for proteomics community. Herein, we developed a user-friendly and powerful stand-alone software, NAguideR, to enable implementation and evaluation of different missing value methods offered by 23 widely used missing-value imputation algorithms. NAguideR further evaluates data imputation results through classic computational criteria and, unprecedentedly, proteomic empirical criteria, such as quant. consistency between different charge-states of the same peptide, different peptides belonging to the same proteins, and individual proteins participating protein complexes and functional interactions. We applied NAguideR into three label-free proteomic datasets featuring peptide-level, protein-level, and phosphoproteomic variables resp., all generated by data independent acquisition mass spectrometry (DIA-MS) with substantial biol. replicates. The results indicate that NAguideR is able to discriminate the optimal imputation methods that are facilitating DIA-MS expts. over those sub-optimal and low-performance algorithms. NAguideR further provides downloadable tables and figures supporting flexible data anal. and interpretation.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3cXis1entrrE&md5=ff8b871b38806a95b7fb0220d5782daf
    24. 24
      Rose, M.; Duhamel, M.; Aboulouard, S.; Kobeissy, F.; Rhun, E. L.; Desmons, A.; Tierny, D.; Fournier, I.; Rodet, F.; Salzet, M. The Role of a Proprotein Convertase Inhibitor in Reactivation of Tumor-Associated Macrophages and Inhibition of Glioma Growth. Mol. Ther.--Oncolytics 2020, 17, 31– 46,  DOI: 10.1016/j.omto.2020.03.005
      [Crossref], [PubMed], [CAS], Google Scholar
      24
      The Role of a Proprotein Convertase Inhibitor in Reactivation of Tumor-Associated Macrophages and Inhibition of Glioma Growth
      Rose, Melanie; Duhamel, Marie; Aboulouard, Soulaimane; Kobeissy, Firas; Le Rhun, Emilie; Desmons, Annie; Tierny, Dominique; Fournier, Isabelle; Rodet, Franck; Salzet, Michel
      Molecular Therapy--Oncolytics (2020), 17 (), 31-46CODEN: MTOHDL; ISSN:2372-7705. (Elsevier Inc.)
      Tumors are characterized by the presence of malignant and non-malignant cells, such as immune cells including macrophages, which are preponderant. Macrophages impact the efficacy of chemotherapy and may lead to drug resistance. In this context and based on our previous work, we investigated the ability to reactivate macrophages by using a proprotein convertases inhibitor. Proprotein convertases process immature proteins into functional proteins, with several of them having a role in immune cell activation and tumorigenesis. Macrophages were treated with a peptidomimetic inhibitor targeting furin, PC1/3, PC4, PACE4, and PC5/6. Their anti-glioma activity was analyzed by mass spectrometry-based proteomics and viability assays in 2D and 3D in vitro cultures. Comparison with temozolomide, the drug used for glioma therapy, established that the inhibitor was more efficient for the redn. of cancer cell d. The inhibitor was also able to reactivate macrophages through the secretion of several immune factors with antitumor properties. Moreover, two proteins considered as good glioma patient survival indicators were also identified in 3D cultures treated with the inhibitor. Finally, we established that the proprotein convertases inhibitor has a dual role as an anti-glioma drug and anti-tumoral macrophage reactivation drug. This strategy could be used together with chemotherapy to increase therapy efficacy in glioma.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3cXht1GntbnL&md5=8198a0c313c461e11bafb3372d34b112
    25. 25
      Cox, J.; Mann, M. MaxQuant enables high peptide identification rates and individualized p.p.b.-range mass accuracies and proteome-wide protein quantification. Nat. Biotechnol. 2008, 26, 1367– 1372,  DOI: 10.1038/nbt.1511
      [Crossref], [PubMed], [CAS], Google Scholar
      25
      MaxQuant enables high peptide identification rates, individualized p.p.b.-range mass accuracies and proteome-wide protein quantification
      Cox, Juergen; Mann, Matthias
      Nature Biotechnology (2008), 26 (12), 1367-1372CODEN: NABIF9; ISSN:1087-0156. (Nature Publishing Group)
      Efficient anal. of very large amts. of raw data for peptide identification and protein quantification is a principal challenge in mass spectrometry (MS)-based proteomics. Here we describe MaxQuant, an integrated suite of algorithms specifically developed for high-resoln., quant. MS data. Using correlation anal. and graph theory, MaxQuant detects peaks, isotope clusters and stable amino acid isotope-labeled (SILAC) peptide pairs as three-dimensional objects in m/z, elution time and signal intensity space. By integrating multiple mass measurements and correcting for linear and nonlinear mass offsets, we achieve mass accuracy in the p.p.b. range, a sixfold increase over std. techniques. We increase the proportion of identified fragmentation spectra to 73% for SILAC peptide pairs via unambiguous assignment of isotope and missed-cleavage state and individual mass precision. MaxQuant automatically quantifies several hundred thousand peptides per SILAC-proteome expt. and allows statistically robust identification and quantification of >4000 proteins in mammalian cell lysates.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BD1cXhsVWjtLzJ&md5=675d31ca84e3a7e4fb9bdd601d8075ea
    26. 26
      O’Brien, J. J.; Gunawardena, H. P.; Paulo, J. A.; Chen, X.; Ibrahim, J. G.; Gygi, S. P.; Qaqish, B. F. The effects of nonignorable missing data on label-free mass spectrometry proteomics experiments. Ann. Appl. Stat. 2018, 12, 2075– 2095,  DOI: 10.1214/18-AOAS1144
      [Crossref], [PubMed], [CAS], Google Scholar
      26
      The effects of nonignorable missing data on label-free mass spectrometry proteomics experiments
      O'Brien Jonathon J; Gunawardena Harsha P; Paulo Joao A; Chen Xian; Ibrahim Joseph G; Gygi Steven P; Qaqish Bahjat F
      The annals of applied statistics (2018), 12 (4), 2075-2095 ISSN:1932-6157.
      An idealized version of a label-free discovery mass spectrometry proteomics experiment would provide absolute abundance measurements for a whole proteome, across varying conditions. Unfortunately, this ideal is not realized. Measurements are made on peptides requiring an inferential step to obtain protein level estimates. The inference is complicated by experimental factors that necessitate relative abundance estimation and result in widespread non-ignorable missing data. Relative abundance on the log scale takes the form of parameter contrasts. In a complete-case analysis, contrast estimates may be biased by missing data and a substantial amount of useful information will often go unused. To avoid problems with missing data, many analysts have turned to single imputation solutions. Unfortunately, these methods often create further difficulties by hiding inestimable contrasts, preventing the recovery of interblock information and failing to account for imputation uncertainty. To mitigate many of the problems caused by missing values, we propose the use of a Bayesian selection model. Our model is tested on simulated data, real data with simulated missing values, and on a ground truth dilution experiment where all of the true relative changes are known. The analysis suggests that our model, compared with various imputation strategies and complete-case analyses, can increase accuracy and provide substantial improvements to interval coverage.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A280%3ADC%252BB3crksFeitA%253D%253D&md5=e7433543cf49d814974145824b08c735
    27. 27
      Marquardt, D. W. Comment - You should standardize the predictor variables in your regression models. J. Am. Stat. Assoc. 1980, 75, 87– 91, 10.1080/01621459.1980.10477430
      Google Scholar
      There is no corresponding record for this reference.
    28. 28
      Menard, S. Standards for standardized logistic regression coefficients. Soc. Forces 2011, 89, 1409– 1428,  DOI: 10.1093/sf/89.4.1409
      [Crossref], Google Scholar
      There is no corresponding record for this reference.
    29. 29
      Kreutz, C. New Concepts for Evaluating the Performance of Computational Methods. IFAC-PapersOnLine 2016, 49, 63– 70,  DOI: 10.1016/j.ifacol.2016.12.104
      [Crossref], Google Scholar
      There is no corresponding record for this reference.
    30. 30
      MATLAB. 9.8.0.1538580 (R2020a); The MathWorks Inc.: Natickand Massachusetts, 2020.
      Google Scholar
      There is no corresponding record for this reference.
    31. 31
      R Core Team. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing. Vienna, Austria, 2021.
      Google Scholar
      There is no corresponding record for this reference.
    32. 32
      Stekhoven, D. J.; Buhlmann, P. MissForest-non-parametric missing value imputation for mixed-type data. Bioinformatics 2012, 28, 112– 118,  DOI: 10.1093/bioinformatics/btr597
      [Crossref], [PubMed], [CAS], Google Scholar
      32
      MissForest-non-parametric missing value imputation for mixed-type data
      Stekhoven, Daniel J.; Buehlmann, Peter
      Bioinformatics (2012), 28 (1), 112-118CODEN: BOINFP; ISSN:1367-4803. (Oxford University Press)
      Modern data acquisition based on high-throughput technol. is often facing the problem of missing data. Algorithms commonly used in the anal. of such large-scale data often depend on a complete set. Missing value imputation offers a soln. to this problem. However, the majority of available imputation methods are restricted to one type of variable only: continuous or categorical. For mixed-type data, the different types are usually handled sep. Therefore, these methods ignore possible relations between variable types. We propose a non-parametric method which can cope with different types of variables simultaneously. Results: We compare several state of the art methods for the imputation of missing values. We propose and evaluate an iterative imputation method (missForest) based on a random forest. By averaging over many unpruned classification or regression trees, random forest intrinsically constitutes a multiple imputation scheme. Using the built-in out-of-bag error ests. of random forest, we are able to est. the imputation error without the need of a test set. Evaluation is performed on multiple datasets coming from a diverse selection of biol. fields with artificially introduced missing values ranging from 10% to 30%. We show that missForest can successfully handle missing values, particularly in datasets including different types of variables. In our comparative study, missForest outperforms other methods of imputation esp. in data settings where complex interactions and non-linear relations are suspected. The out-of-bag imputation error ests. of missForest prove to be adequate in all settings. Addnl., missForest exhibits attractive computational efficiency and can cope with high-dimensional data.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC3MXhs1yms7fO&md5=44a690989dad7424d50a1662f5de2625
    33. 33
      Josse, J.; Husson, F. Handling missing values in exploratory multivariate data analysis methods. J. Soc. Fr. Stat. 2012, 153, 79– 99
      Google Scholar
      There is no corresponding record for this reference.
    34. 34
      Khoonsari, P. E.; Häggmark, A.; Lönnberg, M.; Mikus, M.; Kilander, L.; Lannfelt, L.; Bergquist, J.; Ingelsson, M.; Nilsson, P.; Kultima, K.; Shevchenko, G. Analysis of the Cerebrospinal Fluid Proteome in Alzheimer’s Disease. PLoS One 2016, 11, e0150672  DOI: 10.1371/journal.pone.0150672
      [Crossref], [PubMed], [CAS], Google Scholar
      34
      Analysis of the cerebrospinal fluid proteome in Alzheimer's disease
      Khoonsari, Payam Emami; Haggmark, Anna; Lonnberg, Maria; Mikus, Maria; Kilander, Lena; Lannfelt, Lars; Bergquist, Jonas; Ingelsson, Martin; Nilsson, Peter; Kultima, Kim; Shevchenko, Ganna
      PLoS One (2016), 11 (3), e0150672/1-e0150672/25CODEN: POLNCL; ISSN:1932-6203. (Public Library of Science)
      Alzheimer's disease is a neurodegenerative disorder accounting for more than 50% of cases of dementia. Diagnosis of Alzheimer's disease relies on cognitive tests and anal. of amyloid beta, protein tau, and hyperphosphorylated tau in cerebrospinal fluid. Although these markers provide relatively high sensitivity and specificity for early disease detection, they are not suitable for monitor of disease progression. In the present study, we used label-free shotgun mass spectrometry to analyze the cerebrospinal fluid proteome of Alzheimer's disease patients and non-demented controls to identify potential biomarkers for Alzheimer's disease. We processed the data using five programs (DecyderMS, Maxquant, OpenMS, PEAKS, and Sieve) and compared their results by means of reproducibility and peptide identification, including three different normalization methods. After depletion of high abundant proteins we found that Alzheimer's disease patients had lower fraction of low-abundance proteins in cerebrospinal fluid compared to healthy controls (p<0.05). Consequently, global normalization was found to be less accurate compared to using spiked-in chicken ovalbumin for normalization. In addn., we detd. that Sieve and OpenMS resulted in the highest reproducibility and PEAKS was the programs with the highest identification performance. Finally, we successfully verified significantly lower levels (p<0.05) of eight proteins (A2GL, APOM, C1QB, C1QC, C1S, FBLN3, PTPRZ, and SEZ6) in Alzheimer's disease compared to controls using an antibody-based detection method. These proteins are involved in different biol. roles spanning from cell adhesion and migration, to regulation of the synapse and the immune system.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC28XhtVart7vO&md5=ef30cb763212be43089453d2a5117f7e
    35. 35
      Sturm, M.; Bertsch, A.; Gröpl, C.; Hildebrandt, A.; Hussong, R.; Lange, E.; Pfeifer, N.; Schulz-Trieglaff, O.; Zerck, A.; Reinert, K.; Kohlbacher, O. OpenMS - an open-source software framework for mass spectrometry. BMC Bioinf. 2008, 9, 163  DOI: 10.1186/1471-2105-9-163
      [Crossref], [PubMed], [CAS], Google Scholar
      35
      OpenMS - an open-source software framework for mass spectrometry
      Sturm Marc; Bertsch Andreas; Gropl Clemens; Hildebrandt Andreas; Hussong Rene; Lange Eva; Pfeifer Nico; Schulz-Trieglaff Ole; Zerck Alexandra; Reinert Knut; Kohlbacher Oliver
      BMC bioinformatics (2008), 9 (), 163 ISSN:.
      BACKGROUND: Mass spectrometry is an essential analytical technique for high-throughput analysis in proteomics and metabolomics. The development of new separation techniques, precise mass analyzers and experimental protocols is a very active field of research. This leads to more complex experimental setups yielding ever increasing amounts of data. Consequently, analysis of the data is currently often the bottleneck for experimental studies. Although software tools for many data analysis tasks are available today, they are often hard to combine with each other or not flexible enough to allow for rapid prototyping of a new analysis workflow. RESULTS: We present OpenMS, a software framework for rapid application development in mass spectrometry. OpenMS has been designed to be portable, easy-to-use and robust while offering a rich functionality ranging from basic data structures to sophisticated algorithms for data analysis. This has already been demonstrated in several studies. CONCLUSION: OpenMS is available under the Lesser GNU Public License (LGPL) from the project website at http://www.openms.de.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A280%3ADC%252BD1c3ltF2ktg%253D%253D&md5=99727c3e786a099fdd18786d6e642cd8
    36. 36
      Pursiheimo, A.; Vehmas, A. P.; Afzal, S.; Suomi, T.; Chand, T.; Strauss, L.; Poutanen, M.; Rokka, A.; Corthals, G. L.; Elo, L. L. Optimization of Statistical Methods Impact on Quantitative Proteomics Data. J. Proteome Res. 2015, 14, 4118– 4126,  DOI: 10.1021/acs.jproteome.5b00183
      [ACS Full Text ACS Full Text], [CAS], Google Scholar
      36
      Optimization of Statistical Methods Impact on Quantitative Proteomics Data
      Pursiheimo, Anna; Vehmas, Anni P.; Afzal, Saira; Suomi, Tomi; Chand, Thaman; Strauss, Leena; Poutanen, Matti; Rokka, Anne; Corthals, Garry L.; Elo, Laura L.
      Journal of Proteome Research (2015), 14 (10), 4118-4126CODEN: JPROBS; ISSN:1535-3893. (American Chemical Society)
      As tools for quant. label-free mass spectrometry (MS) rapidly develop, a consensus about the best practices is not apparent. In the work described here, we compared popular statistical methods for detecting differential protein expression from quant. MS data using both controlled expts. with known quant. differences for specific proteins used as stds. as well as "real" expts. where differences in protein abundance are not known a priori. Our results suggest that data-driven reproducibility-optimization can consistently produce reliable differential expression rankings for label-free proteome tools and are straightforward in their application.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2MXhsVWitL%252FI&md5=0c2a241e97545ea49acff53fdcf8c5ec
    37. 37
      Govaert, E.; Van Steendam, K.; Scheerlinck, E.; Vossaert, L.; Meert, P.; Stella, M.; Willems, S.; De Clerck, L.; Dhaenens, M.; Deforce, D. Extracting histones for the specific purpose of label-free MS. Proteomics 2016, 16, 2937– 2944,  DOI: 10.1002/pmic.201600341
      [Crossref], [PubMed], [CAS], Google Scholar
      37
      Extracting histones for the specific purpose of label-free MS
      Govaert, Elisabeth; Van Steendam, Katleen; Scheerlinck, Ellen; Vossaert, Liesbeth; Meert, Paulien; Stella, Martina; Willems, Sander; De Clerck, Laura; Dhaenens, Maarten; Deforce, Dieter
      Proteomics (2016), 16 (23), 2937-2944CODEN: PROTC7; ISSN:1615-9853. (Wiley-VCH Verlag GmbH & Co. KGaA)
      Extg. histones from cells is the first step in studies that aim to characterize histones and their post-translational modifications (hPTMs) with MS. In the last decade, label-free quantification is more frequently being used for MS-based histone characterization. However, many histone extn. protocols were not specifically designed for label-free MS. While label-free quantification has its advantages, it is also very susceptible to tech. variation. Here, we adjust an established histone extn. protocol according to general label-free MS guidelines with a specific focus on minimizing sample handling. These protocols are first evaluated using SDS-PAGE. Hereafter, a selection of extn. protocols was used in a complete histone workflow for label-free MS. All protocols display nearly identical relative quantification of hPTMs. We thus show that, depending on the cell type under investigation and at the cost of some addnl. contaminating proteins, minimizing sample handling can be done during histone isolation. This allows analyzing bigger sample batches, leads to reduced tech. variation and minimizes the chance of in vitro alterations to the hPTM snapshot. Overall, these results allow researchers to det. the best protocol depending on the resources and goal of their specific study. Data are available via ProteomeXchange with identifier PXD002885.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC28XitVSrurzK&md5=682a99c1991f6a9ac6517da0755212a7
    38. 38
      Calf, O. W.; van Dam, N. M.; Weinhold, A.; Huber, H.; Peters, J. L. MTBLS738: Glycoalkaloid composition explains variation in slug resistance in Solanum dulcamara. Oecologia 2018, 187, 495– 506,  DOI: 10.1007/s00442-018-4064-z
      [Crossref], [PubMed], [CAS], Google Scholar
      38
      Glycoalkaloid composition explains variation in slug resistance in Solanum dulcamara
      Calf Onno W; van Dam Nicole M; Huber Heidrun; Peters Janny L; Weinhold Alexander; van Dam Nicole M; van Dam Nicole M
      Oecologia (2018), 187 (2), 495-506 ISSN:.
      In natural environments, plants have to deal with a wide range of different herbivores whose communities vary in time and space. It is believed that the chemical diversity within plant species has mainly arisen from selection pressures exerted by herbivores. So far, the effects of chemical diversity on plant resistance have mostly been assessed for arthropod herbivores. However, also gastropods, such as slugs, can cause extensive damage to plants. Here we investigate to what extent individual Solanum dulcamara plants differ in their resistance to slug herbivory and whether this variation can be explained by differences in secondary metabolites. We performed a series of preference assays using the grey field slug (Deroceras reticulatum) and S. dulcamara accessions from eight geographically distinct populations from the Netherlands. Significant and consistent variation in slug preference was found for individual accessions within and among populations. Metabolomic analyses showed that variation in steroidal glycoalkaloids (GAs) correlated with slug preference; accessions with high GA levels were consistently less damaged by slugs. One, strongly preferred, accession with particularly low GA levels contained high levels of structurally related steroidal compounds. These were conjugated with uronic acid instead of the glycoside moieties common for Solanum GAs. Our results illustrate how intraspecific variation in steroidal glycoside profiles affects resistance to slug feeding. This suggests that also slugs should be considered as important drivers in the co-evolution between plants and herbivores.
      >> More from SciFinder ®
      https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A280%3ADC%252BC1Mvls1Gguw%253D%253D&md5=172dd50c42901c7f65018fa75fab07b3

  • Supporting Information

    Supporting Information

    ARTICLE SECTIONS
    Jump To

    The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.jproteome.1c00119.

    • Sigmoidal decrease of missing values for higher protein intensities (Figure S1); missing value distribution per sample and per protein (Figure S2); distribution of the estimated logistic regression coefficients (Figure S3); DIMA analysis at the peptide level (Figure S4); density plot of the imputed compared to the original data values (Figure S5); principal component analysis before and after imputation (Figure S6); DIMA Implementation (Figure S7); characteristics of the 30 applied imputation algorithms (Table S1); and characteristics of the PRIDE data sets assessed with DIMA (Table S2) (PDF)


    Terms & Conditions

    Most electronic Supporting Information files are available without a subscription to ACS Web Editions. Such files may be downloaded by article for research use (if there is a public use license linked to the relevant article, that license may permit other uses). Permission may be obtained from ACS for other uses through requests via the RightsLink permission system: http://pubs.acs.org/page/copyright/permissions.html.

PDF [ 3MB]

Pair your accounts.

Export articles to Mendeley

Get article recommendations from ACS based on references in your Mendeley library.

Pair your accounts.

Export articles to Mendeley

Get article recommendations from ACS based on references in your Mendeley library.

You’ve supercharged your research process with ACS and Mendeley!

STEP 1:
Click to create an ACS ID

Please note: If you switch to a different device, you may be asked to login again with only your ACS ID.

Please note: If you switch to a different device, you may be asked to login again with only your ACS ID.

Please note: If you switch to a different device, you may be asked to login again with only your ACS ID.

MENDELEY PAIRING EXPIRED
Your Mendeley pairing has expired. Please reconnect