Download Hi-Res ImageDownload to MS-PowerPointCite This:ACS Appl. Mater. Interfaces 2024, 16, 51, 70310-70321

Biological and Medical Applications of Materials and InterfacesDecember 11, 2024

Scalable Accelerated Materials Discovery of Sustainable Polysaccharide-Based Hydrogels by Autonomous Experimentation and Collaborative Learning
Click to copy article linkArticle link copied!

Yang Liu
Yang Liu
Grado Department of Industrial and Systems Engineering, Virginia Tech, Blacksburg, Virginia 24061, United States
Macromolecules Innovation Institute, Virginia Tech, Blacksburg, Virginia 24061, United States
More by Yang Liu
Xubo Yue
Xubo Yue
Department of Mechanical and Industrial Engineering, Northeastern University, Boston, Massachusetts 02115, United States
More by Xubo Yue
Junru Zhang
Junru Zhang
Grado Department of Industrial and Systems Engineering, Virginia Tech, Blacksburg, Virginia 24061, United States
More by Junru Zhang
Zhenghao Zhai
Zhenghao Zhai
Macromolecules Innovation Institute, Virginia Tech, Blacksburg, Virginia 24061, United States
More by Zhenghao Zhai
https://orcid.org/0000-0002-9312-7750
Ali Moammeri
Ali Moammeri
Grado Department of Industrial and Systems Engineering, Virginia Tech, Blacksburg, Virginia 24061, United States
Department of Chemical Engineering, Virginia Tech, Blacksburg, Virginia 24061, United States
More by Ali Moammeri
Kevin J. Edgar
Kevin J. Edgar
Macromolecules Innovation Institute, Virginia Tech, Blacksburg, Virginia 24061, United States
Department of Sustainable Biomaterials, Virginia Tech, Blacksburg, Virginia 24061, United States
More by Kevin J. Edgar
https://orcid.org/0000-0002-9459-9477
Albert S. Berahas
Albert S. Berahas
Department of Industrial and Operations Engineering, University of Michigan, Ann Arbor, Michigan 48109, United States
More by Albert S. Berahas
Raed Al Kontar
Raed Al Kontar
Department of Industrial and Operations Engineering, University of Michigan, Ann Arbor, Michigan 48109, United States
More by Raed Al Kontar
Blake N. Johnson*
Blake N. Johnson
Grado Department of Industrial and Systems Engineering, Virginia Tech, Blacksburg, Virginia 24061, United States
Macromolecules Innovation Institute, Virginia Tech, Blacksburg, Virginia 24061, United States
Department of Chemical Engineering, Virginia Tech, Blacksburg, Virginia 24061, United States
Department of Materials Science and Engineering, Virginia Tech, Blacksburg, Virginia 24061, United States
*Email: [email protected]. Phone: 540-231-0755. Fax: 540-231-3322.
More by Blake N. Johnson
https://orcid.org/0000-0003-4668-2011

Open PDF Supporting Information (2)

ACS Applied Materials & Interfaces

Cite this: ACS Appl. Mater. Interfaces 2024, 16, 51, 70310–70321

Click to copy citationCitation copied!

https://doi.org/10.1021/acsami.4c16614

Published December 11, 2024

CC-BY 4.0 .

Abstract

Click to copy section linkSection link copied!

While some materials can be discovered and engineered using standalone self-driving workflows, coordinating multiple stakeholders and workflows toward a common goal could advance autonomous experimentation (AE) for accelerated materials discovery (AMD). Here, we describe a scalable AMD paradigm based on AE and “collaborative learning”. Collaborative learning using a novel consensus Bayesian optimization (BO) model enabled the rapid discovery of mechanically optimized composite polysaccharide hydrogels. The collaborative workflow outperformed a non-collaborating AMD workflow scaled by independent learning based on the trend of mechanical property evolution over eight experimental iterations, corresponding to a budget limit. After five iterations, four collaborating clients obtained notable material performance (i.e., composition discovery). Collaborative learning by consensus BO can enable scaling and performance optimization for a range of self-driving materials research workflows driven by optimally cooperating humans and machines that share a material design objective.

This publication is licensed under

CC-BY 4.0 .

License Summary*
You are free to share(copy and redistribute) this article in any medium or format and to adapt(remix, transform, and build upon) the material for any purpose, even commercially within the parameters below:
- Creative Commons (CC): This is a Creative Commons license.
- Attribution (BY): Credit must be given to the creator.
View full license
*Disclaimer
This summary highlights only some of the key features and terms of the actual license. It is not a license and has no legal value. Carefully review the actual license before using these materials.
License Summary*
You are free to share(copy and redistribute) this article in any medium or format and to adapt(remix, transform, and build upon) the material for any purpose, even commercially within the parameters below:
- Creative Commons (CC): This is a Creative Commons license.
- Attribution (BY): Credit must be given to the creator.
View full license
*Disclaimer
This summary highlights only some of the key features and terms of the actual license. It is not a license and has no legal value. Carefully review the actual license before using these materials.

Subjects

Keywords

1. Introduction

Click to copy section linkSection link copied!

The Materials Genome Initiative (MGI) aims to achieve rapid and cost-effective materials discovery and engineering. Hybrid materials research infrastructures, including integrated tools, methods, and processes for experimentation, computation, and data analytics, are needed to achieve this goal. (1,2) Automation and machine learning have significantly advanced experimental resources for accelerated materials discovery (AMD), such as establishing workflows based on high-throughput experimentation (HTE) (3,4) and autonomous experimentation (AE). (5,6) While AMD can be achieved by brute force (e.g., HTE) in applications that exhibit a simple design space or large budget, (7) workflow performance and discovery outcomes can be transformed by integrating automation and adaptive sampling methods, such as for AE, in applications that exhibit budget constraints. Given the complexity of functional material composition–process–structure–property relations (i.e., design spaces), adaptive sampling strategies have shown promise to improve several workflow performance metrics, including speed and efficiency, thereby enabling AMD in applications that may exhibit severe budget and resource constraints. (8,9) In particular, applications to AMD of soft materials (e.g., hydrogels) for biomedical applications can exhibit complex design spaces and high experimental cost. While HTE and AMD application to hydrogels is an emerging area (see Table S1), (10−13) applications focused on hard materials have shown value of integrating experimental and computation tools. (2,14)

Bayesian optimization (BO) and active learning use adaptive sampling to achieve a specific learning goal. (15) Active learning, sometimes called “query learning”, is the study of machine learning systems that improve by asking questions. (16) Active learning is instrumental when unlabeled data are numerous, labeling consumes significant resources (e.g., physical or economic), or it is anticipated that many data must be labeled to train a model. (16) There are many ways by which a learner can ask queries. Kusne et al. used Bayesian active learning to discover phase-change memory materials by identifying the composition-phase structure relation over 19 experimental iterations. (17) Min et al. used active learning to discover inorganic materials based on efficient global optimization that satisfy band gap and refractive index constraints over 50 experimental iterations. (18) Oftelie et al. used active learning to design layered materials using BO by identifying optimal structures for band gap and electronic fitness function values. (19)

BO is a sequential process that aims to reduce the number of required experiments by optimizing an unknown black-box (BB) function, (20) such as the true relationship between a material’s composition and properties. BO begins by using a relatively small number of initial data points to learn a surrogate model, often a Gaussian Process (GP) given its intrinsic ability to quantify uncertainty, (21) that approximates the BB function. (22) Subsequently, BO relies on a utility function based on the surrogate to quantify the potential benefits of conducting new experiments (e.g., testing a new formulation). By optimizing the utility function, which balances exploration and exploitation, one can select a new formulation for analysis that will best yield the desired yet unknown composition–process–structure–property relation. Exploration refers to allowing for new feature discovery, while exploitation refers to capitalizing on knowledge already gained. These newly labeled data are then integrated with the existing data set, and the cycle is iterated until available resources (e.g., budget) are utilized or an exit condition is met. Burger et al. implemented BO on a robot to autonomously search for photocatalysts with improved hydrogen production performance. (23) Shields et al. used BO to identify the top-yielding conditions in the Mitsunobu reaction and deoxyfluorination reaction processes. (24)

While adaptive sampling can improve the performance of AMD workflows, such as allowing them to operate under budget constraints, it is prudent to consider its impact on workflow scalability. For example, it may be desirable for multiple clients to leverage their workflows toward a common AMD objective, such as by hierarchical modeling. (25) Advances in the computation and communication power of edge devices (26) now make it plausible for multiple clients, such as humans and machines (e.g., HTE or AE systems), to share information, distribute trial-and-error efforts, and fast-track experimental processes such that all participants benefit, which is a crucial aspect of AMD scalability and performance optimization. Unfortunately, despite its utility in AE and AMD, conventional adaptive sampling methods lack a collaborative element. Thus, there remains a critical need to integrate principles of collaboration with adaptive sampling methods to scale HTE and AE, and AMD applications thereof, using networks of distributed humans and machines.

Here, we demonstrate a methodology for scalable self-driving AMD founded on AE and collaborative learning. We demonstrate the utility and impact of a novel consensus BO framework in which multiple collaborating “clients” locally perform experiments and agree (i.e., reach a consensus) on their individual “next-to-test” formulations to achieve a shared material design objective. This work showcases the potential of merging AE and collaborative machine learning to expedite and scale material discovery initiatives subject to resource and privacy constraints (e.g., budget).

2. Materials and Methods

Click to copy section linkSection link copied!

2.1. Collaborative Learning via Consensus BO

2.1.1. Mathematical Notation

First, we define the mathematical notation used in this work. Assume there are K clients (e.g., experimenters, such as chemists, HTE systems, or AE systems), and each client has a budget of T experiments imposed by resource limitations. Let t ∈ {0,1,···,T – 1} be the iteration index. Consider

D_{k}^{(0)} = {X_{k}^{(0)}, y_{k}^{(0)}}

the initial data set for client k with N_k⁽⁰⁾ samples, where X_k⁽⁰⁾ = (x_k,1,···,x_{k,N_k⁽⁰⁾}) is a matrix that contains the initial input vector data x_k, (i.e., material composition, specifically the concentration of each component in the mixture) and y_k⁽⁰⁾ = (y_k,1,···,y_{k,N_k⁽⁰⁾})^T is a vector that contains the corresponding outputs (i.e., numeric measures of the material properties for each composition, such as the shear storage modulus (G′)). Here, the input data x_k, is of dimension D, where each dimension denotes a component in the material (i.e., D = the number of different species in the mixture). Each client’s goal is to find a formulation (i.e., mixture composition) that optimizes a target material performance or quality measure, specifically G′ in this work. This problem is analogous to a team of cooks attempting to discover a recipe that generates optimal taste. Mathematically, this translates to the following optimization problem:

x_{k}^{*} = argma x_{x} f_{k} (x)

(1)

where f_k is the true BB function each client aims to optimize. To observe f_k(x) using a new formulation x, we need to conduct an experiment and observe a potentially noisy outcome y_k(x). In other words, y_k = f_k + ε_k, where ε_k is an additive noise.

2.1.2. Independent Learning via Traditional BO

In an environment where clients operate independently, they select successive formulations to test by maximizing utility without communicating with others. More specifically, from a mathematical perspective, at each iteration t (i.e., experiment number), a GP surrogate model f̂_k is fit to the data

D_{k}^{(t)}

to estimate f̂_k(x) for any input x. While several kernel functions are applicable for the GP model, the squared exponential kernel function was selected in this study because it defines a smooth function. Importantly, a GP provides a predictive distribution

P_{{\hat{f}}_{k} (x) | D_{k}^{(t)}}

over f̂_k(x). Equipped with the predictive distribution, the k^th client then chooses the next formulation to test by maximizing their expected utility

x_{k}^{(t) new} = a r g m a x_{x} E_{P_{{\hat{f}}_{k} (x) | D_{k}^{(t)}}} [U ({\hat{f}}_{k} (x); D_{k}^{(t)})]

where

U ({\hat{f}}_{k} (x); D_{k}^{(t)})

is a utility function that quantifies the benefit gained if one were to conduct an experiment using a new formulation x, and x_k^(t)new is the recommended next-to-test formulation. Many utility functions have been developed. Among the most used is the expected improvement (EI) utility, (27) which is defined as:

E_{P_{{\hat{f}}_{k} (x) | D_{k}^{(t)}}} [{({\hat{f}}_{k} (x) - y_{k}^{* (t)})}^{+}] = σ_{k}^{(t)} (x; D_{k}^{(t)}) ϕ (z_{k}^{(t)} (x)) + (μ_{k}^{(t)} (x; D_{k}^{(t)}) - y_{k}^{* (t)}) Φ (z_{k}^{(t)} (x))

(2)

where a⁺ = max(a,0), y_k^*(t) is the current best-observed output for client k, ϕ (or Φ) is a probability density function (or cumulative distribution function) of a standard normal random variable,

z_{k}^{(t)} (x) = \frac{μ_{k}^{(t)} (x; D_{k}^{(t)}) - y_{k}^{* (t)}}{σ_{k}^{(t)} (x; D_{k}^{(t)})}

μ_{k}^{(t)} (x; D_{k}^{(t)})

is the predictive mean of the GP at input x, and

σ_{k}^{(t)} (x; D_{k}^{(t)})

is the corresponding variance. The uncertainty from GP prediction (i.e., the variance information σ_k^(t)) enables the utility function to balance exploration and exploitation. EI can be interpreted as finding a new formulation x that has the potential to improve upon the current best-observed output. Surrogate modeling and utility optimization can be done using the GPytorch library. (28)

2.1.3. Collaborative Learning via Consensus BO─Model Description

To foster collaboration and efficiently distribute the workload across multiple clients, we propose a “collaborative BO” approach, where cooperating clients reach a consensus on their next-to-test formulations. Through consensus, clients may borrow information from each other to conduct more efficient and effective tests (e.g., experiments). Our consensus BO approach is accomplished through a consensus matrix W^(t), which regulates the extent to which one client’s decision influences the choices of others. From a mathematical perspective, the new objective function is:

x_{k}^{(t)} = {[x_{C}^{(t)}]}_{k} = a r g m a x_{x} E_{P_{{\hat{f}}_{k} (x) | D_{k}^{(t)}}} [U ({\hat{f}}_{k} (x); D_{k}^{(t)})]

(3)

and

x_{k}^{(t) new} = {[{(W}^{(t)} \otimes I) x_{C}^{(t)}]}_{k}

(4)

where [·]_k represents the k-th block of a vector, x_C^(t) is a vector that concatenates all x_k^(t) for k = {1,···,K}, W^(t) is a dynamic consensus matrix, ⊗ is a Kronecker product operation, and I is an identity matrix. A key property of W^(t) is that it is a symmetric and doubly stochastic matrix (i.e., ∑_kw_kj^(t) = ∑_jw_kj^(t) = 1 for all j, k ∈ {1,···,K}) with non-negative elements.

The consensus BO objective function was solved by the following steps:

1.	At iteration t ∈ {0,1,···,T – 1}, each client k solves $argmax E_{P_{{\hat{f}}_{k} (x) \| D_{k}^{(t)}}} [U ({\hat{f}}_{k} (x); D_{k}^{(t)})]$ and obtains x_k^(t).
2.	Each client k sends its x_k^(t) to a central orchestrator.
3.	The orchestrator concatenates the x_k^(t) of all clients, computes (W^(t) ⊗ I)x_C^(t), and then sends x_k^(t)new to the corresponding client k.
4.	Each client k then conducts a test (e.g., experiment) using the new formulation x_k^(t)new and observes y_k(x_k^(t)new).
5.	Each client k then augments the data set $D_{k}^{(t)}$ by (x_k^(t)new, y_k(x_k^(t)new)) to obtain a new data set $D_{k}^{(t + 1)}$
6.	Each client k then updates its GP surrogate using $D_{k}^{(t + 1)}$ and starts the new iteration t + 1.

The consensus BO framework naturally distributes effort (e.g., experimental workload) across the clients, allowing parallel exploration and exploitation and more efficient experimental design. Additional methodological details of the consensus BO model are provided in our previous work, (29) and our code is provided on GitHub (https://github.com/UMDataScienceLab/Consensus_Bayesian_Opt/tree/main).

A key question that remains is how to design W^(t). To enable effective collaboration, W^(t) is initialized as a uniform matrix whose entries are equal. Mathematically, this is equivalent to:

W^{(0)} = [\begin{array}{l} \frac{1}{K} & \dots & \frac{1}{K} \\ ⋮ & ⋱ & ⋮ \\ \frac{1}{K} & \dots & \frac{1}{K} \end{array}]

(5)

The dynamic consensus matrix is then gradually shrunk to an identity matrix during the collaboration process. For example, if we adjust weights linearly as:

W^{(t + 1)} = W^{(t)} + [\begin{array}{c} \frac{K - 1}{TK} & \dots & - \frac{1}{TK} \\ ⋮ & ⋱ & ⋮ \\ - \frac{1}{TK} & \dots & \frac{K - 1}{TK} \end{array}]

(6)

A finite budget T was assumed in eq 6. Here, T denotes the maximum number of experiments each client can perform using their available resources. When the budget (e.g., quantified as a number of experiments or iterations) is reached, the process stops, and the best design identified is chosen. We emphasize that reaching the budget does not imply that W has reached identity, which only depends on how W decays. Even if this occurs, our model indicates that clients should proceed using W = I, which is essentially independent BO.

An explicit boundary condition constraint can also be incorporated into the utility function (built upon the GP model) and the consensus step. First, when there is a boundary condition constraint, each local client solves the constrained BO problem

a r g \max E_{P_{{\hat{f}}_{k} (x) | D_{k}^{(t)}}} [U ({\hat{f}}_{k} (x); D_{k}^{(t)})] s u c h that c (x) \leq 0

where c(x) ≤ 0 is a constraint set by the user, to ensure that the resulting x_k^(t) lies within the boundary. The constrained function can be optimized using any BO library, such as BoTorch. (30) Second, during the consensus step, the orchestrator concatenates all x_k^(t) and computes (W^(t) ⊗ I)x_C^(t). The resulting x_k^(t)new for each client k naturally satisfies the boundary constraint since it is the convex combination of solutions {x_k^(t)}_k. As for the case study, the constraints were set to be the range of the input parameters.

In summary, each client, indexed by k, first initially optimizes its individual utility function to determine a candidate experimental formulation, which is stored in a vector x_k^(t). These formulation vectors are then shared with a central orchestrator and concatenated into a vector x_C^(t) = (x₁^(t)T, ···, x_K^(t)T)^T. Subsequently, the consensus matrix W^(t) aggregates information from all clients. This operation results in a new vector x^(t)new = (x₁^(t)newT, ···, x_K^(t)newT)^T that contains personalized next-to-test formulations for each client. The orchestrator then shares these next-to-test formulations with the clients for testing. This process is repeated until an exit condition is met, such as reaching a budget limit.

The consensus framework has several notable features. First, it is clear from the consensus BO objective function (i.e., eqs 3 and 4) that W^(t) plays a critical role in determining the next-to-test formulation for each client k (i.e., x_k^(t)new). W^(t) controls the extent to which each client affects the decisions of others. As a result, W^(t) adds flexibility to the optimization problem and allows a group of clients to determine the degree of collaboration dynamically (i.e., with iteration number). Second, when W^(t) is an identity matrix I, the consensus step will keep the original vector x_C^(t) unchanged (i.e., (W^(t) ⊗ I)x_C^(t) = (I ⊗ I)x_C^(t) = x_C^(t)). This is equivalent to no information sharing; therefore, the consensus BO model becomes traditional BO. Third, since W^(t) is doubly stochastic, this ensures that the formulations of all clients will stay in the solution space that has been explored. Finally, our consensus BO objective function generates an individualized solution for each client that allows the group of clients to distribute effort (e.g., experimental or computational), allowing parallel exploration–exploitation and more efficient experimental design. The intuition behind this design is as follows. In the early stages, a given client k may not have enough observations to obtain a high-quality surrogate model and, therefore, needs to borrow information from other clients. In the late stages, as client k accumulates sufficient data and can construct a high-quality local surrogate model, it will focus more on its local problem to find a client-specific optimal formulation. This distributed active learning approach with consensus BO transforms traditional BO into a new collaborative paradigm, which can accelerate the optimal design process by effectively distributing experimental efforts across a network of clients.

2.2. Materials

Acetic acid (AcOH) was from Sigma-Aldrich. Multi-reducing end alginate (M-alginate) and carboxymethyl chitosan sodium salt (CMCS) were synthesized as previously reported. (31,32) Lead zirconate titanate (PZT-5A, 72.4 × 72.4 × 0.127 mm³ wafer) with nickel (Ni) electrodes was from Piezo Systems (Woburn, MA). Glass cylinders and ethanol (200 proof) were from Fisher Scientific. Polyurethane (Fast-Drying) and Loctite EA 1C-LV epoxy adhesive were from Minwax and Henkel, respectively. Deionized water (DIW; 18.2 MΩ) was from a commercial system (Direct-Q 3 UV Water Purification Systems; Millipore).

2.3. High-Throughput Synthesis

Material libraries of unique composition that spanned the material design space, specifically the formulation space, were fabricated in a 96-well plate using a previously reported automated HTE system based on robotic dispensing and sensing. (7) The precursor solutions were dispensed from a syringe (55 or 10 mL) using a straight cylindrical nozzle (32 or 27 gauge, respectively). Samples in the first library were formulated with 0.75 wt % M-alginate and 0.75 wt % CMCS precursor solutions using identical dispensing times. The total concentration of M-alginate and CMCS in each sample was a constant 0.66 wt % (i.e., 0.33 wt % M-alginate and 0.33 wt % CMCS). After dispensing, each well’s precursor solutions were robotically mixed for 30 s using a helical stirring tool. The mass of AcOH dispensed across all wells ranged from 0.02 to 52 mg. Varying but known quantities of DIW were then dispensed across all samples to make the total mass of each sample identical (390 mg). The composite polysaccharide hydrogels were then obtained by allowing the cross-linking reaction to proceed for 12 h in a closed, humid environment at room temperature. The experiment was thrice performed.

Samples in a second library were formulated with 0.75 wt % M-alginate and 0.75 wt % CMCS precursor solutions with dispensing time per well ranging from 0 to 23.5 s and 23.5 to 0 s, respectively, using a step size of 0.5 s. M-alginate and CMCS precursor solutions were dispensed in opposite paths. The total concentration of M-alginate and CMCS in each sample (i.e., well) ranged from 0–0.66 wt %, but the total solids concentration of each sample (i.e., the sum of the M-alginate and CMCS weight percentages) was a constant 0.66 wt % (see Supporting Data Set). After dispensing, each well’s precursor solutions were robotically mixed for 30 s using a helical stirring tool. Varying but known volumes of AcOH was then dispensed across the wells. Three concentrations of AcOH were used depending on the given library to maximize the explored design space. The dispensing time of AcOH (17.4 M) ranged from 0.8 to 10 s across the plate with a step size of 0.4 s. The dispensing time of 8.3 and 0.69 wt % AcOH solutions ranged from 0.5 to 6 s with a step size of 0.5 s. Varying but known quantities of DIW were then dispensed across all samples to make the total mass of each sample identical (390 mg). The composite polysaccharide hydrogels were formed as described for the first library. Three sample replicates were prepared and simultaneously characterized for each formulation.

2.4. High-Throughput Characterization

Hydrogel viscoelastic properties, specifically the shear storage modulus (G′), were measured using a previously reported well plate-compatible high-throughput characterization (HTC) system. (33) The dynamic-mode cantilever rheometer fabrication method and measurement principle can be found in our previous reports. (33−36) Measurements were obtained using a 60 s dwell time per sample to generate a steady-state sensor response. The sensor stabilization time between samples was 60 s. Following analysis of the last sample, the rheometer signal was continuously monitored to verify a stable baseline. The mechanical property of each sample, specifically G′, is correlated with the cantilever rheometer phase angle at resonance (φ) and was calculated based on the steady-state values of φ and resonant frequency (f₀) obtained during the dwell period using the fluid–structure interaction model of Mather et al. (37) The associated shear rate is 2πf₀.

2.5. Scalable Self-Driving AMD via AE and Collaborative Learning

HTC of initial material libraries formulated by high-throughput synthesis identified the feasible input domain for the collaborative and non-collaborative (i.e., independent) learning models. Specifically, the labeled data generated from the initial sample library identified the feasible range for each mixture component. These labeled data, which established constraints on the design space, were also used to generate the first-to-test formulations for each client in the collaborative learning and control groups for iteration 0. Five points per client were sampled from these initial labeled data based on Latin hypercube design, (38) which yielded the first formulation to test for each client based on the learning models. One additional input point per iteration was added to each client’s data set to determine the next formulation to test.

The samples were formulated sequentially adding the M-alginate and CMCS precursor solutions followed by DIW in a 96-well plate. The sol mixture was then homogenized using a digital plate shaker (LSE digital microplate shaker; Corning) at 700 rpm for 30 min. AcOH was then added to make the total mass of each sample identical to those of the initial sample libraries, and the samples were mixed for 5 min at 700 rpm. The samples were then cross-linked as described in Section 2.3. Following hydrogel formation, the samples were characterized as described in Section 2.4, resulting in new labeled data. The newly labeled data were then incorporated into the learning models as described in Section 2.1, which generated the next-to-test formulations for all clients for the subsequent iteration. This process was repeated until an imposed budget was met.

2.6. Statistical Analysis

The reported results correspond to the mean and standard deviation of all measurements obtained from triplicate samples. A Student’s t test was used to characterize the significance level of the difference between the optimal target property obtained using collaborative vs independent learning.

3. Results and Discussion

Click to copy section linkSection link copied!

3.1. Initializing Scalable Self-Driving AMD via AE and Collaborative Learning Using Labeled Data Acquired by HTE

As shown in Figure 1A, this work aimed to generate a scalable self-driving AMD workflow founded on AE and collaborative learning. We focus on a glycomaterial engineering application, given the opportunities associated with sustainability and the design challenges posed by the complexity of polysaccharide chemistry. Alginate and chitosan are important renewable glycomaterials that are widely used as scaffolds for tissue engineering and drug release applications. (39−44) Alginate exhibits biocompatibility and tunable mechanical properties by controlling polymer content and cross-linking chemistry. (45) Chitosan offers attractive mechanical properties and reactive groups, such as for applications requiring robust mechanical properties and chemical alteration. (46) M-alginate is a multi-reducing end polysaccharide and forms hydrogels with CMCS. As shown in Figure 1B,C, we focus on a shared design objective among four clients who may work independently or collaboratively to discover the optimal formulation that maximizes the hydrogel mechanical properties, specifically G′, through a scaled AMD workflow.

Figure 1. (A) Illustration of the scalable accelerated materials discovery (AMD) workflow driven by autonomous experimentation (AE) and collaborative learning. A novel collaborative learning model guides the selection of formulations to be tested by collaborating clients using labeled data generated by high-throughput experimentation (HTE). (B) Flowchart for scalable AMD via AE and collaborative learning based on consensus BO using a network of clients. (C) Illustration of the difference between traditional non-collaborative BO (i.e., independent learning) and collaborative BO with consensus (i.e., collaborative learning). Created in BioRender. Liu, Y. (2024) https://BioRender.com/r26y741.

The reaction schemes to synthesize M-alginate and CMCS are shown in Figure 2A. While aqueous solutions of M-alginate and CMCS form hydrogels (see Figure 2B), it remains a challenge to discover optimal composite polysaccharide hydrogels that best combine the advantages of the individual polysaccharides due to the large formulation space and challenges of blending glycomaterials. In addition, the currently unknown gelation mechanism of M-alginate-CMCS hydrogels further motivates the use of HTE or AE to interrogate the material design space.

Figure 2. (A) Reaction schemes for synthesis of multi-reducing end alginate (M-alginate) and carboxymethyl chitosan sodium salt (CMCS). (B) Illustration and photographs of acid-catalyzed composite polysaccharide hydrogel cross-linking (gelation) in the presence of acetic acid (AcOH) (photographs: left = sol; right = gel; sample composition: M-alginate 0.37 wt %, CMCS 0.37 wt %, AcOH 1.72 wt %, DIW 97.54 wt %).

While considerable strides have been made in AMD applications to hard and brittle materials research, relatively less progress has been made in soft materials research applications, such as hydrogels, due to the challenges of adapting conventional methods for formulation, synthesis, and characterization of soft materials to autonomous or high-throughput formats (47,48) and the high dimensionality of the material design spaces. For example, while solutions of M-alginate and CMCS can form composite hydrogels in the presence of AcOH (see Figure 2B), (49,50) the resultant mechanical properties are highly dependent on the material formulation (i.e., four-component mixture), which contains M-alginate, CMCS, AcOH, and water. Thus, acid-catalyzed composite M-alginate-CMCS hydrogels exhibit a complex design space, which presents a challenge to AMD using brute-force HTE.

The scalable AMD workflow driven by AE and collaborative learning is shown in Figure 1, which consists of an AE workflow for material synthesis and characterization driven by a group of clients with a shared material design objective. Representative impedance spectra of the cantilever rheometer are shown in Figure 3A before and after gelation for a single formulation. As shown in Figure S1, while the spectra of three different representative formulations before gelation were similar, although the mixtures exhibited significantly different compositions (sol 1: 0.25 wt % M-alginate and 0.41 wt % CMCS; sol 2: 0.29 wt % M-alginate and 0.37 wt % CMCS; sol 3: 0.42 wt % M-alginate and 0.24 wt %, CMCS), the spectra after gelation by AcOH addition (sol 1: 5.19 wt % AcOH; sol 2: 7.74 wt % AcOH; sol 3: 6.24 wt % AcOH) were significantly different. This result demonstrates the impact of the formulation on the composite hydrogel G′.

Figure 3. (A) Cantilever rheometer impedance spectra: (1) before measurement (air), (2) upon submersion in an M-alginate-CMCS precursor solution (sol), and (3) after gelation by reaction with AcOH for 12 h (gel). (B) Results of the HTE study that established the constraints on AcOH concentration used for M-alginate-CMCS composite hydrogel cross-linking. (C) Heat map showing the amount of AcOH dispensed in each sample of an initial 48-sample M-alginate-CMCS composite hydrogel library. (D) Real-time phase angle at resonance (φ(t)) response (i.e., raw HTC data) for the library described in (C). (E) Heat map of the steady-state φ response for the library described in (C) (12 samples were removed (i.e., row D) due to observed phase separation; thus, row D in the heat map does not contain sample information). (F) The composition–property relation for the M-alginate-CMCS composite hydrogel associated with the data in (E) presented as a ternary diagram (the DIW content of each sample is provided as Supporting Information).

Given the complex design space associated with the composition–property relation of the acid-catalyzed composite M-alginate-CMCS hydrogel, we first examined the dependence of the sol–gel transition on the AcOH concentration by HTE using a 48-sample library in which the polymer content of all samples was constant (0.33 wt % M-alginate and 0.33 wt % CMCS). The total mass of each sample was 390 mg. As shown in Figure 3B, the relationship between φ and the AcOH concentration exhibited a sigmoidal trend. These data provided the upper and lower bounds (i.e., constraints) of AcOH content for the design space (48 and 0.02 mg per well equivalent to 12.31 and ∼5 × 10^–5 wt %, respectively).

We next determined the design space constraints for the M-alginate and CMCS components by HTE using a second 48-sample library in which all samples exhibited an identical total polymer mass of 0.66 wt % (i.e., wt %_M-alginate + wt %_CMCS = 0.66 wt %) but the ratio of the macromolecular components was varied (see Figure 3C). Similarly, the total mass of each sample was 390 mg. The first and last samples in the library exhibited the lowest and highest relative amounts of M-alginate, respectively. The first and last samples in the library exhibited the highest and lowest relative amounts of CMCS, respectively. The AcOH content was randomly varied within the previously determined range to maximize the explored design space. As shown in Figure S2, the first 12 samples (i.e., composite polysaccharide hydrogels), which were rich in CMCS, exhibited phase separation and inhomogeneity. Thus, these compositions were removed from the design space. Figure 3D shows the raw HTC data for the other 36 samples, which formed homogeneous hydrogels. The corresponding steady-state value of φ for the 36 samples obtained by HTC is shown as a heat map in Figure 3E. The random distribution of φ across the material library and the absence of a sigmoidal trend observed in Figure 3B were expected results caused by the addition of AcOH in random volumes across the library. The corresponding composition–property relation data for the acid-catalyzed composite M-alginate-CMCS hydrogel is shown in Figure 3F as a ternary diagram and highlights the hydrogel G′ dependence on the formulation, specifically the AcOH content. We note that while the DIW content of each sample is not shown in Figure 3F, given the challenge of visualizing the composition–property relation for the four-component mixture, the DIW content of each sample can be obtained from knowledge of the total sample mass, which was identical for all samples. The DIW content of all samples shown in Figure 3F is provided as Supporting Information (see Supporting Data Set).

Relatively more labeled data were generated near the lower left corner of the diagram by design, given that the mass fraction of AcOH, which exhibited a maximum of 12.13 wt %, was higher than that of the polymer components, which exhibited a maximum of 0.66 wt %. The data in Figure 3F suggest that an imbalance in the mixture’s two macromolecular components resulted in relatively weak hydrogels. Increasing the AcOH mass fraction appeared to increase the composite hydrogel G′ regardless of the M-alginate:CMCS ratio. In summary, the two initial HTE studies generated the following material design constraints for the learning algorithms: 0.16 < wt %_M-Alginate < 0.66, 0.01 < wt %_CMCS < 0.45, 0.01 < wt %_AcOH < 10, and wt %_M-Alginate + wt %_CMCS = 0.66.

3.2. Scalable AMD via AE and Collaborative Learning with Consensus BO

Although brute-force HTE provides a methodology for exploring the design space and offers the potential for material discovery and optimization, it may be prohibitive in many applications from a resource expenditure perspective (i.e., physical or economic resource constraints), such as glycomaterial engineering applications. Also, there is no assurance that the formulations that exhibit promising performance or quality metrics are optimal. Alternatively, we assert that AE, in combination with collaborative learning, can establish a scalable self-driving AMD workflow capable of efficiently discovering an optimal design (i.e., formulation) within practical budget constraints. Thus, we used the labeled data generated by the initial HTE studies and the generated design constraints to efficiently explore the material design space by collaborative learning with consensus BO.

As shown in the left panel of Figure 1C, the methodology was validated by considering four clients that share an AMD objective, specifically, the aim to rapidly optimize G′ of an acid-catalyzed composite polysaccharide hydrogel composed of M-alginate and CMCS subject to an imposed budget. As shown in the right panel of Figure 1C, a control group was also examined based on independent learning (i.e., traditional BO) in which clients did not collaborate. Thus, for each iteration of the collaborative AMD workflow, the consensus BO model generated four different formulations to test (i.e., one per client). For example, given that all samples have identical total mass the four-component hydrogel can be framed as a three-component material design problem. Thus, D = 3 given the identical mass of all samples removes one degree of freedom from the material design problem and K = 4 given we consider a network of four collaborating clients. If we consider clients use the following consensus matrix

W^{(t)} = [\begin{array}{l} 0.6 & 0.2 & 0.1 & 0.1 \\ 0.2 & 0.5 & 0.2 & 0.1 \\ 0.1 & 0.2 & 0.5 & 0.1 \\ 0.1 & 0.1 & 0.1 & 0.7 \end{array}]

then, if the local utility maximizers of the clients are x_C^(t) = ([1, 2, 1], [1, 2, 1], [2, 3, 3], [2, 1, 3])^T, we obtain (W^(t) ⊗ I)x_C^(t) = (1.2, 2, 1.4, 1.3, 2.1, 1.6, 1.5, 2.2, 2.1, 1.8, 1.4, 2.6), where x₁^(t)new = (1.2, 2, 1.4)^T for client 1, x₂^(t)new = (1.3, 2.1, 1.6)^T for client 2, x₃^(t)new = (1.5, 2.2, 2.1)^T for client 3, and x₄^(t)new = (1.8, 1.4, 2.6)^T for client 4. The independent learning model also contained four clients, each testing one formulation per iteration (we remind the reader that this group considers W = I in the consensus BO model).

To generate each client’s first formulation to test, we randomly sampled 20 of 36 labeled data points in Figure 3F without replacement and assigned them randomly to the four clients in each group. The learning models were initialized using five points per client since BO, and active learning in general, relies on a surrogate (a GP) to help inform the next data point to obtain, and such surrogates need at least three to six data points. Each client tested their formulation in triplicate to establish the error of the data label (i.e., the variance of G′ for a given formulation). HTC data for a single iteration involving 24 samples (8 clients × 1 sample/client × 3 replicates = 24 samples with eight different compositions, one unique composition per client), is shown in Figure S3 as a representative case. In many practical AMD applications, resource constraints limit the number of possible experimental iterations. Our budget constraints in this study restricted the number of iterations to eight. The number of data points used by each client and their respective learning model increased by one after each iteration, regardless of the group. Thus, consensus does not increase the number of data points used by the collaborative group relative to the control group; instead, it helps the collaborative group choose the next experiment better. We designed consensus in this way for two fundamental reasons: (1) privacy: sharing responses across clients may not respect privacy (e.g., sharing outcomes of medical treatment), and (2) heterogeneity: if clients have heterogeneity, then one client’s input–output tuple cannot be used in another client’s surrogate, as their BB functions exhibit differences.

3.3. Comparison of Scalable Self-Driving AMD Performance via Collaborative vs Independent AE

The results of the collaborative learning-driven AMD workflow over eight iterations are shown in Figure 4 compared to the control group. Figure 4 displays the progress of the collaborative workflow in terms of the cantilever rheometer raw data and the associated G′. For client 1, the trend is increasing from iterations 1 to 4 (see Figure 4B,D). After iteration 4, the consensus BO algorithm exploits the current best solution or explores other promising regions, reflecting the exploration–exploitation nature of BO. This explains why the trend does not continue to increase steadily. The overall trend indicates that the consensus BO model has discovered a desirable solution after four iterations. Similarly, the overall trend fluctuates for other clients but generally increases, with the best result also discovered after several iterations. Despite the limited number of iterations, which we remind the reader is determined by the imposed budget limit, collaboration among four clients working toward the shared material design objective achieved superior performance relative to the non-collaborative clients within six iterations (see Figure 4A,C). As shown in Figure 4E, the formulations identified by the four collaborating clients across iterations 6–8 exhibited an average absolute phase angle value (|φ|) of 86.17 ± 0.18° and an average G′ of 91.6 ± 15.3 kPa. The average changes in average |φ| and G′ for successive iterations (across iterations 6–8) were 0.25 ± 0.18° and 27.1 ± 15.3 kPa, respectively. In contrast, the average |φ| and G′ across iterations 6–8 in the independent learning group were relatively smaller and exhibited larger variance (85.96 ± 0.28° and 81.2 ± 15.9 kPa, respectively). Likewise, the average changes in average |φ| and G′ for successive iterations (across iterations 6–8) and the associated variances were relatively larger in the control group (0.42 ± 0.28° and 34.0 ± 15.9 kPa, respectively). Figure 4E also shows a comparison of the mechanical properties associated with the optimal formulation identified by collaborative vs independent learning. The optimal formulation was identified in iteration 4 by client 1 in the collaborative learning group and iteration 8 by client 1 in the independent learning group.

Figure 4. Absolute phase angle at resonance (|φ|) (A, B) and G′ (C, D) for scalable AMD driven by AE and collaborative learning with consensus BO vs traditional active learning by BO (i.e., independent learning). Labeled data acquired using brute-force HTE were used to inform the selection of formulations for iteration 0. (E) The mean G′ from iteration 6 to 8 and the optimal G′ obtained from collaborative learning and independent learning (* indicates p < 0.05). (F) Composition–property (G′) relation for the composite hydrogel generated by collaborative vs independent learning in terms of a ternary diagram (the DIW content of each sample is provided as Supporting Information).

Figure 4 demonstrates that the scaled, self-driving AMD workflow involving collaborating clients exhibited a more consistent trend in the target property to be optimized (G′) compared to the group that contained the same number of non-collaborating clients. We remind the reader that the storage modulus E′ can be obtained from G′ and the Poisson ratio, typically assumed to be 0.5 in hydrogels. (51) In conclusion, the higher average G′ and E′ obtained by cooperating clients (p = 0.014) in the final budgeted iterations indicates the AMD scalability and improved performance offered by collaborative learning. Given the time required for each iteration was the same in both groups, the collaborative learning method also increased the efficiency of the scaled AMD workflow.

The trends in formulation discovery visualized as a composition–property relation are highlighted in Figure 4F for both groups. The DIW content of all samples is provided in Supporting Information. The ternary diagram highlights the data distribution that the collaborating and non-collaborating clients labeled through their search for an optimal formulation that produces a maximum hydrogel G′ and E′ (the initial labeled data obtained by brute-force HTE are excluded for clarity of data visualization). The collaborative learning model incentivized the cooperating clients to focus on exploration in the design space where G′ was relatively high. In contrast, the independent learning model drove the noninteracting clients to explore a larger design space where G′ exhibited a wide range. This is an expected result based on the design of the consensus BO model. For example, the consensus matrix (see eqs 5 and 6) is designed to support clients during the early stages when they lack sufficient data to fit high-quality surrogates, allowing them to leverage information from each other. In the later stages (i.e., iterations), we reduced the influence of different clients by shrinking the off-diagonal elements in the consensus matrix to enable clients to focus on their individual experiments and obtain client-specific optimal formulations.

While the insight into the physiochemical origins (e.g., bonding and structure) that describe a material’s given profile of physical or material properties is valuable, a detailed investigation of the composition–structure–property relation remains challenging. For example, given the complexity of polysaccharide chemistry, such as arising from the multi-reducing end nature of M-alginate used in this study, a detailed investigation of material structure is nontrivial and beyond the scope of this work. Hydrogels are bulk materials in which the local structures and interactions may vary. For example, the local structure inside polymer networks exhibits various potential forms, such as strands, physical entanglement, and chemical cross-linking. (52)

Despite its success in this case study, the current collaborative learning model has some limitations. First, the current model is limited to applications that involve single-objective optimization problems. Our ongoing work focuses on learning strategies for multiobjective optimization applications, such as the accelerated discovery of materials with several optimized performance and quality characteristics. Second, while applying the methodology to a three-dimensional design space is meaningful and establishes proof of principle, higher dimensionality is necessary for many other applications. For example, optimization of material processing parameters can also be included in the input space, given their potential impact on material structure and properties. Finally, the current collaborative learning model lacks explainability. For example, while adaptive sampling strategies can identify optimal formulations, additional work is required postdiscovery to understand the underlying chemistry and engineering principles that yield the process–composition–structure–property relation. Additional constraints guided by domain knowledge in chemistry and engineering can also potentially be integrated with the model to improve performance and explainability.

4. Conclusions

Click to copy section linkSection link copied!

This study addressed a grand challenge in AE and AMD: scalability. While brute-force HTE can accomplish AMD for some materials, such as those with relatively simple design spaces and large budgets, the designer or manufacturer lacks certainty (e.g., quantitative assurance) that the identified formulation is optimal in the given design space, which is particularly challenging for complex functional materials, and an imposed budget will be optimally expended. We first demonstrated that self-driving AMD workflows can be scaled by AE and collaborative learning (i.e., “collaborative AE”), which distributes effort across multiple clients working toward a shared material design optimization objective. We showed that collaborative learning via consensus BO, which is based on cooperation among multiple clients, can establish a scalable self-driving AMD workflow that outperforms that driven by traditional independent learning using BO. The methodology was applied to AMD of novel composite polysaccharide hydrogels that exhibit optimal mechanical properties, specifically G′ and E′. This work demonstrates the capability of using consensus BO to scale AMD workflows using groups of interacting humans and machines.

Data Availability

Click to copy section linkSection link copied!

An open-source software implementation of consensus Bayesian optimization is available at https://github.com/UMDataScienceLab/Consensus_Bayesian_Opt/tree/main.

Supporting Information

Click to copy section linkSection link copied!

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsami.4c16614.

Additional methodological details of the fluid–structure interaction model; experimental results related to HTE studies, and a description of sample compositions, and Cantilever spectra; hydrogel photographs; raw sensor data, and summary of recent progress in literature (PDF)
Representative data set for AE (XLSX)

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

Click to copy section linkSection link copied!

Corresponding Author
- Blake N. Johnson - Grado Department of Industrial and Systems Engineering, Virginia Tech, Blacksburg, Virginia 24061, United States; Macromolecules Innovation Institute, Virginia Tech, Blacksburg, Virginia 24061, United States; Department of Chemical Engineering, Virginia Tech, Blacksburg, Virginia 24061, United States; Department of Materials Science and Engineering, Virginia Tech, Blacksburg, Virginia 24061, United States; https://orcid.org/0000-0003-4668-2011; Email: [email protected]
Authors
- Yang Liu - Grado Department of Industrial and Systems Engineering, Virginia Tech, Blacksburg, Virginia 24061, United States; Macromolecules Innovation Institute, Virginia Tech, Blacksburg, Virginia 24061, United States
- Xubo Yue - Department of Mechanical and Industrial Engineering, Northeastern University, Boston, Massachusetts 02115, United States
- Junru Zhang - Grado Department of Industrial and Systems Engineering, Virginia Tech, Blacksburg, Virginia 24061, United States
- Zhenghao Zhai - Macromolecules Innovation Institute, Virginia Tech, Blacksburg, Virginia 24061, United States; https://orcid.org/0000-0002-9312-7750
- Ali Moammeri - Grado Department of Industrial and Systems Engineering, Virginia Tech, Blacksburg, Virginia 24061, United States; Department of Chemical Engineering, Virginia Tech, Blacksburg, Virginia 24061, United States
- Kevin J. Edgar - Macromolecules Innovation Institute, Virginia Tech, Blacksburg, Virginia 24061, United States; Department of Sustainable Biomaterials, Virginia Tech, Blacksburg, Virginia 24061, United States; https://orcid.org/0000-0002-9459-9477
- Albert S. Berahas - Department of Industrial and Operations Engineering, University of Michigan, Ann Arbor, Michigan 48109, United States
- Raed Al Kontar - Department of Industrial and Operations Engineering, University of Michigan, Ann Arbor, Michigan 48109, United States
Author Contributions
Conceptualization: B.N.J., R.A.K., Y.L., and X.Y. Polymer Synthesis: Z.Z. and K.E. Model Formulation: R.A.K., X.Y., and A.S.B. Experimentation: Y.L., J.Z., and Z.Z. Modeling: XY, R.A.K., Y.L., and B.N.J. Data Analysis: B.N.J., Y.L., A.M., R.A.K., and X.Y. Draft Preparation and Revision: Y.L., A.M., B.N.J., R.A.K., and X.Y.
Notes
The authors declare no competing financial interest.

Acknowledgments

Click to copy section linkSection link copied!

This work was supported by GlycoMIP, a National Science Foundation (NSF) Materials Innovation Platform funded through Cooperative Agreement DMR-1933525. This work was also supported by NSF CMMI-2144147 (R.A.K.), NSF CBET-2126176 (B.N.J.), and NSF CBET- 2141008 (B.N.J.). The authors acknowledge the use of Biorender software in preparation of Figure 1 (Created in BioRender. Liu, Y. (2024) https://BioRender.com/r26y741) and the Table of Contents (ToC) (Created in BioRender. Liu, Y. (2024) https://BioRender.com/e97r830).

References

Click to copy section linkSection link copied!

This article references 52 other publications.

1
Green, M. L.; Choi, C. L.; Hattrick-Simpers, J. R.; Joshi, A. M.; Takeuchi, I.; Barron, S. C.; Campo, E.; Chiang, T.; Empedocles, S.; Gregoire, J. M. Fulfilling The Promise of The Materials Genome Initiative With High-Throughput Experimental Methodologies. Appl. Phys. Rev. 2017, 4 (1), 011105 DOI: 10.1063/1.4977487
Google Scholar
1
Fulfilling the promise of the materials genome initiative with high-throughput experimental methodologies
Green, M. L.; Choi, C. L.; Hattrick-Simpers, J. R.; Joshi, A. M.; Takeuchi, I.; Barron, S. C.; Campo, E.; Chiang, T.; Empedocles, S.; Gregoire, J. M.; Kusne, A. G.; Martin, J.; Mehta, A.; Persson, K.; Trautt, Z.; Van Duren, J.; Zakutayev, A.
Applied Physics Reviews (2017), 4 (1), 011105/1-011105/18CODEN: APRPG5; ISSN:1931-9401. (American Institute of Physics)
The Materials Genome Initiative, a national effort to introduce new materials into the market faster and at lower cost, has made significant progress in computational simulation and modeling of materials. To build on this progress, a large amt. of exptl. data for validating these models, and informing more sophisticated ones, will be required. High-throughput experimentation generates large vols. of exptl. data using combinatorial materials synthesis and rapid measurement techniques, making it an ideal exptl. complement to bring the Materials Genome Initiative vision to fruition. This paper reviews the state-of-the-art results, opportunities, and challenges in high-throughput experimentation for materials design. A major conclusion is that an effort to deploy a federated network of high-throughput exptl. (synthesis and characterization) tools, which are integrated with a modern materials data infrastructure, is needed. (c) 2017 American Institute of Physics.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2sXltlKkurY%253D&md5=8b20fd6d59f2ef776b47e8e8a4503651
2
de Pablo, J. J.; Jackson, N. E.; Webb, M. A.; Chen, L.-Q.; Moore, J. E.; Morgan, D.; Jacobs, R.; Pollock, T.; Schlom, D. G.; Toberer, E. S. New frontiers for the materials genome initiative. npj Comput. Mater. 2019, 5 (1), 41 DOI: 10.1038/s41524-019-0173-4
Google Scholar
There is no corresponding record for this reference.
3
Maier, W. F.; Stöwe, K.; Sieg, S. Combinatorial and High-Throughput Materials Science. Angew. Chem., Int. Ed. 2007, 46 (32), 6016– 6067, DOI: 10.1002/anie.200603675
Google Scholar
3
Combinatorial and high-throughput materials science
Maier, Wilhelm F.; Stoewe, Klaus; Sieg, Simone
Angewandte Chemie, International Edition (2007), 46 (32), 6016-6067, S6016/1-S6016/4CODEN: ACIEF5; ISSN:1433-7851. (Wiley-VCH Verlag GmbH & Co. KGaA)
A review. High-through-put techniques are already used routinely for research on complex materials, such as polymers, electronic materials, and catalysts. Remarkable achievements are related to parallel syntheses and analyses, data-mining technologies, modeling approaches, as well as evolutionary strategies for materials development and formulations. The discoveries and success stories document the power of these emerging technologies.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BD2sXpsFyit7s%253D&md5=bd0bde3ae47b664b7a15120522d09b05
4
Eyke, N. S.; Koscher, B. A.; Jensen, K. F. Toward Machine Learning-Enhanced High-Throughput Experimentation. Trends Chem. 2021, 3 (2), 120– 132, DOI: 10.1016/j.trechm.2020.12.001
Google Scholar
4
Toward Machine Learning-Enhanced High-Throughput Experimentation
Eyke, Natalie S.; Koscher, Brent A.; Jensen, Klavs F.
Trends in Chemistry (2021), 3 (2), 120-132CODEN: TCRHBQ; ISSN:2589-5974. (Cell Press)
A review. Recent literature suggests that the fields of machine learning (ML) and high-throughput experimentation (HTE) have sep. received considerable attention from chemists and engineers, leading to the development of powerful reactivity models and platforms capable of rapidly performing thousands of reactions. The merger of ML with HTE presents a wealth of opportunities for the exploration of chem. space, but the integration of the two has yet to be fully realized. We highlight examples of recent developments in ML and HTE that collectively suggest the utility of their integration. Our anal. highlights the complementarity of the two fields, while exposing a no. of obstacles that can and should be overcome to take full advantage of this merger and thereby accelerate chem. research.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3MXhtFeltrnO&md5=231220ca97d31091f9d83a7a246fcb27
5
Häse, F.; Roch, L. M.; Aspuru-Guzik, A. Next-Generation Experimentation with Self-Driving Laboratories. Trends Chem. 2019, 1 (3), 282– 291, DOI: 10.1016/j.trechm.2019.02.007
Google Scholar
5
Next-Generation Experimentation with Self-Driving Laboratories
Hase, Florian; Roch, Loic M.; Aspuru-Guzik, Alan
Trends in Chemistry (2019), 1 (3), 282-291CODEN: TCRHBQ; ISSN:2589-5974. (Cell Press)
A review. The ever-growing demand for advanced functional materials requires disruption of conventional approaches to experimentation and acceleration of the discovery process. State-of-the-art approaches to scientific discovery are inherently slow, capital intensive, and have arguably reached a plateau. Significant advances are possible when rethinking and redesigning the traditional experimentation process. Self-driving labs. promise to substantially accelerate the discovery process by augmenting automated experimentation platforms with artificial intelligence (AI). AI methods actively search for promising exptl. procedures by hypothesizing about their outcomes based on previous expts. This feedback loop is crucial to reduce the no. of expts. needed for discovery. Supplying automated platforms with AI enables self-driving labs. to fully embrace the vision of autonomous experimentation.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3cXitFehtL3L&md5=8118dcb8ae47482044b08d47d0f43952
6
Abolhasani, M.; Kumacheva, E. The rise of self-driving labs in chemical and materials sciences. Nat. Synth. 2023, 2 (6), 483– 492, DOI: 10.1038/s44160-022-00231-0
Google Scholar
There is no corresponding record for this reference.
7
Liu, Y.; Zhang, J.; Zhang, Y.; Yoon, H. Y.; Jia, X.; Roman, M.; Johnson, B. N. Accelerated Engineering of Optimized Functional Composite Hydrogels via High-Throughput Experimentation. ACS Appl. Mater. Interfaces 2023, 15 (45), 52908– 52920, DOI: 10.1021/acsami.3c11483
Google Scholar
There is no corresponding record for this reference.
8
Park, T.; Kim, E.; Sun, J.; Kim, M.; Hong, E.; Min, K. Rapid discovery of promising materials via active learning with multi-objective optimization. Mater. Today Commun. 2023, 37, 107245 DOI: 10.1016/j.mtcomm.2023.107245
Google Scholar
There is no corresponding record for this reference.
9
Bai, Y.; Khoo, Z. H. J.; Made, R. I.; Xie, H.; Lim, C. Y. J.; Handoko, A. D.; Chellappan, V.; Cheng, J. J.; Wei, F.; Lim, Y.-F. Closed Loop Multi-Objective Optimization for Cu-Sb-S Photoelectrocatalytic Materials Discovery. Adv. Mater. 2023, 36 (2), 2304269 DOI: 10.1002/adma.202304269
Google Scholar
There is no corresponding record for this reference.
10
Orlova, T.; Piven, A.; Darmoroz, D.; Aliev, T.; Razik, T.; Boitsev, A.; Grafeeva, N.; Skorb, E. Machine learning for soft and liquid molecular materials. Digital Discovery 2023, 2 (2), 298– 315, DOI: 10.1039/D2DD00132B
Google Scholar
There is no corresponding record for this reference.
11
Li, Z. H.; Song, P. R.; Li, G. F.; Han, Y. F.; Ren, X. X.; Bai, L.; Su, J. C. AI energized hydrogel design, optimization and application in biomedicine. Mater. Today Bio 2024, 25, 101014 DOI: 10.1016/j.mtbio.2024.101014
Google Scholar
There is no corresponding record for this reference.
12
Oliveira, M. B.; Mano, J. F. High-throughput screening for integrative biomaterials design: exploring advances and new trends. Trends Biotechnol. 2014, 32 (12), 627– 636, DOI: 10.1016/j.tibtech.2014.09.009
Google Scholar
There is no corresponding record for this reference.
13
Callahan, L. S. Combinatorial Method/High Throughput Strategies for Hydrogel Optimization in Tissue Engineering Applications. Gels 2016, 2 (2), 18 DOI: 10.3390/gels2020018
Google Scholar
There is no corresponding record for this reference.
14
Green, M. L.; Choi, C. L.; Hattrick-Simpers, J. R.; Joshi, A. M.; Takeuchi, I.; Barron, S. C.; Campo, E.; Chiang, T.; Empedocles, S.; Gregoire, J. M. Fulfilling the promise of the materials genome initiative with high-throughput experimental methodologies. Appl. Phys. Rev. 2017, 4 (1), 011105 DOI: 10.1063/1.4977487
Google Scholar
14
Fulfilling the promise of the materials genome initiative with high-throughput experimental methodologies
Green, M. L.; Choi, C. L.; Hattrick-Simpers, J. R.; Joshi, A. M.; Takeuchi, I.; Barron, S. C.; Campo, E.; Chiang, T.; Empedocles, S.; Gregoire, J. M.; Kusne, A. G.; Martin, J.; Mehta, A.; Persson, K.; Trautt, Z.; Van Duren, J.; Zakutayev, A.
Applied Physics Reviews (2017), 4 (1), 011105/1-011105/18CODEN: APRPG5; ISSN:1931-9401. (American Institute of Physics)
The Materials Genome Initiative, a national effort to introduce new materials into the market faster and at lower cost, has made significant progress in computational simulation and modeling of materials. To build on this progress, a large amt. of exptl. data for validating these models, and informing more sophisticated ones, will be required. High-throughput experimentation generates large vols. of exptl. data using combinatorial materials synthesis and rapid measurement techniques, making it an ideal exptl. complement to bring the Materials Genome Initiative vision to fruition. This paper reviews the state-of-the-art results, opportunities, and challenges in high-throughput experimentation for materials design. A major conclusion is that an effort to deploy a federated network of high-throughput exptl. (synthesis and characterization) tools, which are integrated with a modern materials data infrastructure, is needed. (c) 2017 American Institute of Physics.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2sXltlKkurY%253D&md5=8b20fd6d59f2ef776b47e8e8a4503651
15
Di Fiore, F.; Nardelli, M.; Mainini, L. Active Learning and Bayesian Optimization: A Unified Perspective to Learn with a Goal. Arch. Comput. Methods Eng. 2024, 31, 2985– 3013, DOI: 10.1007/s11831-024-10064-z
Google Scholar
There is no corresponding record for this reference.
16
Settles, B. Active Learning; Morgan & Claypool Publishers, 2012.
Google Scholar
There is no corresponding record for this reference.
17
Kusne, A. G.; Yu, H.; Wu, C.; Zhang, H.; Hattrick-Simpers, J.; DeCost, B.; Sarker, S.; Oses, C.; Toher, C.; Curtarolo, S. On-the-fly closed-loop materials discovery via Bayesian active learning. Nat. Commun. 2020, 11 (1), 5966 DOI: 10.1038/s41467-020-19597-w
Google Scholar
17
On-the-fly closed-loop materials discovery via Bayesian active learning
Kusne, A. Gilad; Yu, Heshan; Wu, Changming; Zhang, Huairuo; Hattrick-Simpers, Jason; DeCost, Brian; Sarker, Suchismita; Oses, Corey; Toher, Cormac; Curtarolo, Stefano; Davydov, Albert V.; Agarwal, Ritesh; Bendersky, Leonid A.; Li, Mo; Mehta, Apurva; Takeuchi, Ichiro
Nature Communications (2020), 11 (1), 5966CODEN: NCAOBW; ISSN:2041-1723. (Nature Research)
Active learning-the field of machine learning (ML) dedicated to optimal expt. design-has played a part in science as far back as the 18th century when Laplace used it to guide his discovery of celestial mechanics. In this work, we focus a closed-loop, active learning-driven autonomous system on another major challenge, the discovery of advanced materials against the exceedingly complex synthesis-processes-structure-property landscape. We demonstrate an autonomous materials discovery methodol. for functional inorg. compds. which allow scientists to fail smarter, learn faster, and spend less resources in their studies, while simultaneously improving trust in scientific results and machine learning tools. This robot science enables science-over-the-network, reducing the economic impact of scientists being phys. sepd. from their labs. The real-time closed-loop, autonomous system for materials exploration and optimization (CAMEO) is implemented at the synchrotron beamline to accelerate the interconnected tasks of phase mapping and property optimization, with each cycle taking seconds to minutes. We also demonstrate an embodiment of human-machine interaction, where human-in-the-loop is called to play a contributing role within each cycle. This work has resulted in the discovery of a novel epitaxial nanocomposite phase-change memory material.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3cXisVOnsL3N&md5=138fb31700fa99c1962c048fe7829145
18
Min, K.; Cho, E. Accelerated Discovery of Novel Inorganic Materials with Desired Properties Using Active Learning. J. Phys. Chem. C 2020, 124 (27), 14759– 14767, DOI: 10.1021/acs.jpcc.0c00545
Google Scholar
18
Accelerated Discovery of Novel Inorganic Materials with Desired Properties Using Active Learning
Min, Kyoungmin; Cho, Eunseog
Journal of Physical Chemistry C (2020), 124 (27), 14759-14767CODEN: JPCCCK; ISSN:1932-7447. (American Chemical Society)
Construction of prediction models using machine learning algorithms on existing databases expands the search limit of undiscovered structures, in principle, to the entire materials space. However, because of uncertainties in machine learning prediction, the suggested properties are not always promising; thus, improving the database quality is mandatory for validation as well as improvement in prediction accuracy. To achieve this, we herein implement an active learning process, beginning with a limited no. of databases, to find materials satisfying target properties (band gap and refractive index) with minimized trials and errors. The regression model is initially trained with only around 2% of the entire search space, and 20 new databases, suggested from the optimization schemes, are added at each optimization process. Between exploration, exploitation, random selection, and the Bayesian optimization method, the Bayesian method exhibits the best performance in finding the no. of materials that satisfies the criteria within limited trials In addn., the structure with the max. target property values is found after searching only around 7.0% and 7.7% of the entire database for band gap and refractive index, resp. Current results clearly confirm that the active learning process can be accelerated to find ideal materials satisfying target properties with minimized resources.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3cXht1Git77J&md5=92537da74ec4d8a89d711934b7a91114
19
Oftelie, L. B.; Rajak, P.; Kalia, R. K.; Nakano, A.; Sha, F.; Sun, J.; Singh, D. J.; Aykol, M.; Huck, P.; Persson, K.; Vashishta, P. Active learning for accelerated design of layered materials. npj Comput. Mater. 2018, 4 (1), 74 DOI: 10.1038/s41524-018-0129-0
Google Scholar
There is no corresponding record for this reference.
20
Shahriari, B.; Swersky, K.; Wang, Z.; Adams, R. P.; Freitas, N. d. Taking the Human Out of the Loop: A Review of Bayesian Optimization. Proc. IEEE 2016, 104 (1), 148– 175, DOI: 10.1109/jproc.2015.2494218
Google Scholar
There is no corresponding record for this reference.
21
Chen, H.; Zheng, L.; Kontar, R. A.; Raskutti, G. Gaussian process parameter estimation using mini-batch stochastic gradient descent: convergence guarantees and empirical benefits. J. Mach. Learn. Res. 2022, 23 (1), 1– 59
Google Scholar
There is no corresponding record for this reference.
22
Rasmussen, C. E.; Williams, C. K. I. Gaussian Processes for Machine Learning; The MIT Press, 2005.
Google Scholar
There is no corresponding record for this reference.
23
Burger, B.; Maffettone, P. M.; Gusev, V. V.; Aitchison, C. M.; Bai, Y.; Wang, X.; Li, X.; Alston, B. M.; Li, B.; Clowes, R. A mobile robotic chemist. Nature 2020, 583 (7815), 237– 241, DOI: 10.1038/s41586-020-2442-2
Google Scholar
23
A mobile robotic chemist
Burger, Benjamin; Maffettone, Phillip M.; Gusev, Vladimir V.; Aitchison, Catherine M.; Bai, Yang; Wang, Xiaoyan; Li, Xiaobo; Alston, Ben M.; Li, Buyi; Clowes, Rob; Rankin, Nicola; Harris, Brandon; Sprick, Reiner Sebastian; Cooper, Andrew I.
Nature (London, United Kingdom) (2020), 583 (7815), 237-241CODEN: NATUAS; ISSN:0028-0836. (Nature Research)
Abstr.: Technologies such as batteries, biomaterials and heterogeneous catalysts have functions that are defined by mixts. of mol. and mesoscale components. As yet, this multi-length-scale complexity cannot be fully captured by atomistic simulations, and the design of such materials from first principles is still rare1-5. Likewise, exptl. complexity scales exponentially with the no. of variables, restricting most searches to narrow areas of materials space. Robots can assist in exptl. searches6-14 but their widespread adoption in materials research is challenging because of the diversity of sample types, operations, instruments and measurements required. Here we use a mobile robot to search for improved photocatalysts for hydrogen prodn. from water15. The robot operated autonomously over eight days, performing 688 expts. within a ten-variable exptl. space, driven by a batched Bayesian search algorithm16-18. This autonomous search identified photocatalyst mixts. that were six times more active than the initial formulations, selecting beneficial components and deselecting neg. ones. Our strategy uses a dexterous19,20 free-roaming robot21-24, automating the researcher rather than the instruments. This modular approach could be deployed in conventional labs. for a range of research problems beyond photocatalysis.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3cXhtlCqur%252FN&md5=6a13f9399b94cca78cdeac312a2e4563
24
Shields, B. J.; Stevens, J.; Li, J.; Parasram, M.; Damani, F.; Alvarado, J. I. M.; Janey, J. M.; Adams, R. P.; Doyle, A. G. Bayesian reaction optimization as a tool for chemical synthesis. Nature 2021, 590 (7844), 89– 96, DOI: 10.1038/s41586-021-03213-y
Google Scholar
24
Bayesian reaction optimization as a tool for chemical synthesis
Shields, Benjamin J.; Stevens, Jason; Li, Jun; Parasram, Marvin; Damani, Farhan; Alvarado, Jesus I. Martinez; Janey, Jacob M.; Adams, Ryan P.; Doyle, Abigail G.
Nature (London, United Kingdom) (2021), 590 (7844), 89-96CODEN: NATUAS; ISSN:0028-0836. (Nature Research)
Abstr.: Reaction optimization is fundamental to synthetic chem., from optimizing the yield of industrial processes to selecting conditions for the prepn. of medicinal candidates1. Likewise, parameter optimization is omnipresent in artificial intelligence, from tuning virtual personal assistants to training social media and product recommendation systems2. Owing to the high cost assocd. with carrying out expts., scientists in both areas set numerous (hyper)parameter values by evaluating only a small subset of the possible configurations. Bayesian optimization, an iterative response surface-based global optimization algorithm, has demonstrated exceptional performance in the tuning of machine learning models3. Bayesian optimization has also been recently applied in chem.4-9; however, its application and assessment for reaction optimization in synthetic chem. has not been investigated. Here we report the development of a framework for Bayesian reaction optimization and an open-source software tool that allows chemists to easily integrate state-of-the-art optimization algorithms into their everyday lab. practices. We collect a large benchmark dataset for a palladium-catalyzed direct arylation reaction, perform a systematic study of Bayesian optimization compared to human decision-making in reaction optimization, and apply Bayesian optimization to two real-world optimization efforts (Mitsunobu and deoxyfluorination reactions). Benchmarking is accomplished via an online game that links the decisions made by expert chemists and engineers to real expts. run in the lab. Our findings demonstrate that Bayesian optimization outperforms human decisionmaking in both av. optimization efficiency (no. of expts.) and consistency (variance of outcome against initially available data). Overall, our studies suggest that adopting Bayesian optimization methods into everyday lab. practices could facilitate more efficient synthesis of functional chems. by enabling better-informed, data-driven decisions about which expts. to run.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3MXjt1SjtLg%253D&md5=4ae6e5eb6ae548b86f651b4a72508e28
25
Kusne, A. G.; McDannald, A. Scalable multi-agent lab framework for lab optimization. Matter 2023, 6 (6), 1880– 1893, DOI: 10.1016/j.matt.2023.03.022
Google Scholar
There is no corresponding record for this reference.
26
Kontar, R.; Shi, N. C.; Yue, X. B.; Chung, S.; Byon, E.; Chowdhury, M.; Jin, J. H.; Kontar, W.; Masoud, N.; Nouiehed, M. The Internet of Federated Things (IoFT). IEEE Access 2021, 9, 156071– 156113, DOI: 10.1109/ACCESS.2021.3127448
Google Scholar
There is no corresponding record for this reference.
27
Frazier, P. I. A. Tutorial on Bayesian optimization, arXiv:1807.02811. arXiv.org e-Print archive, 2018. https://arXiv.org/abs/1807.02811.
Google Scholar
There is no corresponding record for this reference.
28
Gardner, J. R.; Pleiss, G.; Bindel, D.; Weinberger, K. Q.; Wilson, A. G. GPyTorch: Blackbox Matrix-Matrix Gaussian Process Inference with GPU Acceleration, Advances in Neural Information Processing Systems; NeurIPS, 2018.
Google Scholar
There is no corresponding record for this reference.
29
Yue, X.; Al Kontar, R.; Berahas, A. S.; Liu, Y.; Zai, Z.; Edgar, K.; Johnson, B. N. Collaborative and Distributed Bayesian Optimization via Consensus: Showcasing the Power of Collaboration for Optimal Design, arXiv:2306.14348. arXiv.org e-Print archive, 2023. https://arXiv.org/abs/2306.14348.
Google Scholar
There is no corresponding record for this reference.
30
Balandat, M.; Karrer, B.; Jiang, D. R.; Daulton, S.; Letham, B.; Wilson, A. G.; Bakshy, E. BOTORCH: A Framework for Efficient Monte-Carlo Bayesian Optimization, Advances in Neural Information Processing Systems; NeurIPS, 2020.
Google Scholar
There is no corresponding record for this reference.
31
Zhai, Z.; Zhou, Y.; Korovich, A. G.; Hall, B. A.; Yoon, H. Y.; Yao, Y.; Zhang, J.; Bortner, M. J.; Roman, M.; Madsen, L. A.; Edgar, K. J. Synthesis and Characterization of Multi-Reducing-End Polysaccharides. Biomacromolecules 2023, 24 (6), 2596– 2605, DOI: 10.1021/acs.biomac.3c00104
Google Scholar
There is no corresponding record for this reference.
32
Zhou, Y.; Zhai, Z.; Yao, Y.; Stant, J. C.; Landrum, S. L.; Bortner, M. J.; Frazier, C. E.; Edgar, K. J. Oxidized hydroxypropyl cellulose/carboxymethyl chitosan hydrogels permit pH-responsive, targeted drug release. Carbohydr. Polym. 2023, 300, 120213 DOI: 10.1016/j.carbpol.2022.120213
Google Scholar
32
Oxidized hydroxypropyl cellulose/carboxymethyl chitosan hydrogels permit pH-responsive, targeted drug release
Zhou, Yang; Zhai, Zhenghao; Yao, Yimin; Stant, John C.; Landrum, Sarah L.; Bortner, Michael J.; Frazier, Charles E.; Edgar, Kevin J.
Carbohydrate Polymers (2023), 300 (), 120213CODEN: CAPOD8; ISSN:0144-8617. (Elsevier Ltd.)
Polysaccharide-based Schiff base hydrogels have promise for drug delivery, tissue engineering, and many other applications due to their reversible imine bond crosslinks. We describe herein pH-responsive, injectable, and self-healing hydrogels prepd. by reacting oxidized hydroxypropyl cellulose (Ox-HPC) with carboxymethyl chitosan (CMCS). Simple combination of ketones from Ox-HPC side chains with amines from CMCS in water provides a dynamic, hydrophilic polysaccharide network. The reversible nature of these imine bonds in the presence of water provides a hydrogel with injectable and self-healing properties. Phenylalanine as a model amine-contg. drug was linked by imine bonds to Ox-HPC within the hydrogel. Phenylalanine release was faster at the pH of the extracellular space around tumors (6.8) than in normal tissues (7.4), a surprising degree of pH sensitivity. Therefore, Ox-HPC/CMCS hydrogels show promise as drug carriers that may selectively target even slightly lower pH environments like the extracellular milieu around cancer cells.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB38Xisleisb%252FP&md5=f93efc00c7e1e21a973a0925d701a6d8
33
Zhang, J.; Liu, Y.; Sekhar P, D. C.; Singh, M.; Tong, Y.; Kucukdeger, E.; Yoon, H. Y.; Haring, A. P.; Roman, M.; Kong, Z.; Johnson, B. N. Rapid, autonomous high-throughput characterization of hydrogel rheological properties via automated sensing and physics-guided machine learning. Appl. Mater. Today 2023, 30, 101720 DOI: 10.1016/j.apmt.2022.101720
Google Scholar
There is no corresponding record for this reference.
34
Haring, A. P.; Singh, M.; Koh, M.; Cesewski, E.; Dillard, D. A.; Kong, Z. J.; Johnson, B. N. Real-time characterization of hydrogel viscoelastic properties and sol-gel phase transitions using cantilever sensors. J. Rheol. 2020, 64 (4), 837– 850, DOI: 10.1122/8.0000009
Google Scholar
34
Real-time characterization of hydrogel viscoelastic properties and sol-gel phase transitions using cantilever sensors
Haring, Alexander P.; Singh, Manjot; Koh, Miharu; Cesewski, Ellen; Dillard, David A.; Kong, Zhenyu "James"; Johnson, Blake N.
Journal of Rheology (Melville, NY, United States) (2020), 64 (4), 837-850CODEN: JORHD2; ISSN:0148-6055. (American Institute of Physics)
Here, we report for the first time that resonance in dynamic-mode cantilever sensors persists in hydrogels and enables the real-time characterization of hydrogel viscoelastic properties and the continuous monitoring of sol-gel phase transitions (i.e., gelation and dissoln. processes). Real-time tracking of piezoelec.-excited millimeter cantilever (PEMC) sensor resonant frequency (fair = 55.4 ± 8.8 kHz; n = 5 sensors) and quality factor (Q; Qair = 23.8 ± 1.5) enabled continuous monitoring of high-frequency hydrogel shear storage and loss moduli (G'f and G"f, resp.) calcd. by sensor data and fluid-structure interaction models. Changes in the sensor phase angle, quality factor, and high-frequency shear moduli obtained at the resonant frequency (G'fand G"f) correlated with low-frequency moduli obtained at 1 Hz using dynamic mech. anal. Characterization studies were performed using phys. and chem. crosslinked hydrogel systems, including gelatin hydrogels (6-10 wt.%) and alginate hydrogels (0.25-0.75 wt.%). The sensor exhibited a dynamic range from the rheol. properties of inviscid solns. to hydrogels with high-frequency moduli of 80 kPa and low-frequency moduli of 26 kPa. The sensor exhibited a limit of detection of 260 Pa and 1.9 kPa for changes in hydrogel storage modulus (E') based on the sensor's phase angle and quality factor responses, resp. We also show that sensor data enable quant. characterization of gelation process dynamics using a modified Hill model. This work suggests that cantilever sensors provide a promising platform for the sensor-based characterization of hydrogels, such as quantification of viscoelastic properties and real-time monitoring of gelation processes. (c) 2020 American Institute of Physics.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3cXps1Ohurw%253D&md5=a263d7613603732f7e688f0ca9d15282
35
Singh, M.; Zhang, J.; Bethel, K.; Liu, Y.; Davis, E. M.; Zeng, H.; Kong, Z.; Johnson, B. N. Closed-Loop Controlled Photopolymerization of Hydrogels. ACS Appl. Mater. Interfaces 2021, 13 (34), 40365– 40378, DOI: 10.1021/acsami.1c11779
Google Scholar
35
Closed-Loop Controlled Photopolymerization of Hydrogels
Singh, Manjot; Zhang, Junru; Bethel, Keturah; Liu, Yang; Davis, Eric M.; Zeng, Haibo; Kong, Zhenyu; Johnson, Blake N.
ACS Applied Materials & Interfaces (2021), 13 (34), 40365-40378CODEN: AAMICK; ISSN:1944-8244. (American Chemical Society)
Here, we present a closed-loop controlled photopolymn. process for fabrication of hydrogels with controlled storage moduli. Hydrogel crosslinking was assocd. with a significant change in the phase angle of a piezoelec. cantilever sensor and established the timescale of the photopolymn. process. The compn., structure, and mech. properties of the fabricated hydrogels were characterized using Raman spectroscopy, SEM (SEM), and dynamic mech. anal. (DMA). We found that the storage moduli of photocured poly(ethylene glycol) dimethacrylate (PEGDMA) and poly(N-isopropylacrylamide) (PNIPAm) hydrogels could be controlled using bang-bang and fuzzy logic controllers. Bang-bang controlled photopolymn. resulted in const. overshoot of the storage modulus setpoint for PEGDMA hydrogels, which was mitigated by setpoint correction and fuzzy logic control. SEM and DMA studies showed that the network structure and storage modulus of PEGDMA hydrogels were dependent on the cure time and temporal profile of UV exposure during photopolymn. This work provides an advance in pulsed and continuous photopolymn. processes for hydrogel engineering based on closed-loop control that enables reproducible fabrication of hydrogels with controlled mech. properties.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3MXhvVart7fO&md5=1522b64d1d8f3d22093bf2c407f8ede0
36
Liu, Y.; Bethel, K.; Singh, M.; Zhang, J.; Ashkar, R.; Davis, E. M.; Johnson, B. N. Comparison of Bulk- vs Layer-by-Layer-Cured Stimuli-Responsive PNIPAM–Alginate Hydrogel Dynamic Viscoelastic Property Response via Embedded Sensors. ACS Appl. Polym. Mater. 2022, 4 (8), 5596– 5607, DOI: 10.1021/acsapm.2c00634
Google Scholar
There is no corresponding record for this reference.
37
Mather, M. L.; Rides, M.; Allen, C. R. G.; Tomlins, P. E. Liquid Viscoelasticity Probed by a Mesoscale Piezoelectric Bimorph Cantilever. J. Rheol. 2012, 56 (1), 99– 112, DOI: 10.1122/1.3670732
Google Scholar
37
Liquid viscoelasticity probed by a mesoscale piezoelectric bimorph cantilever
Mather, Melissa L.; Rides, Martin; Allen, Crispin R. G.; Tomlins, Paul E.
Journal of Rheology (Melville, NY, United States) (2012), 56 (1), 99-112CODEN: JORHD2; ISSN:0148-6055. (American Institute of Physics)
The viscoelastic properties of fluids are key to their performance in industries ranging from biotechnol. to the automotive industry. Traditionally, fluid viscoelastic properties are monitored with rheometers but these are expensive, require a skilled operator, function over a relatively limited frequency range and are not suitable for in situ monitoring. Piezoelec. cantilevers capable of in situ assessment of the rheol. properties of relatively small fluid vols. have the potential to overcome many of these limitations and can be fabricated into low cost probes. Rheol. assessment of test fluids using piezoelec. cantilevers is typically made through anal. of the cantilever's resonant oscillation in the fluids. For accurate results, the damping of the cantilever should be low as quantified by a high quality factor Q. This can be difficult in fluids of high viscosity particularly for microscopic cantilevers. In this paper, a "mesoscale" piezoelec. bimorph cantilever was used. The mesoscale refers to a size regime intermediate between microscopic and macroscopic, in this work the cantilever used has dimensions of the order of millimeters. This mesoscale cantilever displayed a sufficiently high Q to probe the rheol. properties of highly damping and elastic fluids in situ. The developed probe will be ideally suited to in-process monitoring of high value products such as those in the biotechnol. industry. (c) 2012 American Institute of Physics.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC38XntF2rtQ%253D%253D&md5=609f9c33b3c1cc15bc68d73d6e927421
38
Wei-Liem, L. On Latin hypercube sampling. Ann. Stat. 1996, 24 (5), 2058– 2080, DOI: 10.1214/aos/1069362310
Google Scholar
There is no corresponding record for this reference.
39
Yang, Q.; Peng, J.; Xiao, H.; Xu, X.; Qian, Z. Polysaccharide hydrogels: Functionalization, construction and served as scaffold for tissue engineering. Carbohydr. Polym. 2022, 278, 118952 DOI: 10.1016/j.carbpol.2021.118952
Google Scholar
39
Polysaccharide hydrogels: Functionalization, construction and served as scaffold for tissue engineering
Yang, Qian; Peng, Jinrong; Xiao, Haitao; Xu, Xuewen; Qian, Zhiyong
Carbohydrate Polymers (2022), 278 (), 118952CODEN: CAPOD8; ISSN:0144-8617. (Elsevier Ltd.)
A review. Polysaccharide hydrogels have been widely utilized in tissue engineering. They interact with the organismal environments, modulating the cargos release and realizing of long-term survival and activations of living cells. In this review, the potential strategies for modification of polysaccharides were introduced firstly. It is not only used to functionalize the polysaccharides for the consequent formation of hydrogels, but also used to introduce versatile side groups for the regulation of cell behavior. Then, techniques and underlying mechanisms in inducing the formation of hydrogels by polysaccharides or their derivs. are briefly summarized. Finally, the applications of polysaccharide hydrogels in vivo, mainly focus on the performance for alleviation of foreign-body response (FBR) and as cell scaffolds for tissue regeneration, are exemplified. In addn., the perspectives and challenges for further research are addressed. It aims to provide a comprehensive framework about the potentials and challenges that the polysaccharide hydrogels confronting in tissue engineering.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3MXislelsbvL&md5=301eb9a47feab5504afaa492b72872ea
40
Zhang, M.; Ma, H.; Wang, X.; Yu, B.; Cong, H.; Shen, Y. Polysaccharide-based nanocarriers for efficient transvascular drug delivery. J. Controlled Release 2023, 354, 167– 187, DOI: 10.1016/j.jconrel.2022.12.051
Google Scholar
40
Polysaccharide-based nanocarriers for efficient transvascular drug delivery
Zhang, Min; Ma, He; Wang, Xijie; Yu, Bing; Cong, Hailin; Shen, Youqing
Journal of Controlled Release (2023), 354 (), 167-187CODEN: JCREEC; ISSN:0168-3659. (Elsevier B.V.)
A review. Polysaccharide-based nanocarriers (PBNs) are the focus of extensive investigation because of their biocompatibility, low cost, wide availability, and chem. versatility, which allow a wide range of anticancer agents to be loaded within the nanocarriers. Similar to other nanocarriers, most PBNs are designed to extravasate out of tumor vessels, depending on the enhanced permeability and retention (EPR) effect. However, the EPR effect is compromised in some tumors due to the heterogeneity of tumor structures. Transvascular transport efficacy is decreased by complex blood vessels and condensed tumor stroma. The limited extravasation impedes efficient drug delivery into tumor parenchyma, and thus affects the subsequent tumor accumulation, which hinders the therapeutic effect of PBNs. Therefore, overcoming the biol. barriers that restrict extravasation from tumor vessels is of great importance in PBN design. Many strategies have been developed to enhance the EPR effect that involve nanocarrier property regulation and tumor structure remodeling. Moreover, some researchers have proposed active transcytosis pathways that are complementary to the paracellular EPR effect to increase the transvascular extravasation efficiency of PBNs. In this review, we summarize the recent advances in the design of PBNs with enhanced transvascular transport to enable optimization of PBNs in the extravasation of the drug delivery process. We also discuss the obstacles and challenges that need to be addressed to clarify the transendothemial mechanism of PBNs and the potential interactions between extravasation and other drug delivery steps.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3sXnsl2ntQ%253D%253D&md5=17215726f9de1cfb4c60de184cb94bda
41
Kim, H.-L.; Jung, G.-Y.; Yoon, J.-H.; Han, J.-S.; Park, Y.-J.; Kim, D.-G.; Zhang, M.; Kim, D.-J. Preparation and characterization of nano-sized hydroxyapatite/alginate/chitosan composite scaffolds for bone tissue engineering. Mater. Sci. Eng.: C 2015, 54, 20– 25, DOI: 10.1016/j.msec.2015.04.033
Google Scholar
41
Preparation and characterization of nano-sized hydroxyapatite/alginate/chitosan composite scaffolds for bone tissue engineering
Kim, Hye-Lee; Jung, Gil-Yong; Yoon, Jun-Ho; Han, Jung-Suk; Park, Yoon-Jeong; Kim, Do-Gyoon; Zhang, Miqin; Kim, Dae-Joon
Materials Science & Engineering, C: Materials for Biological Applications (2015), 54 (), 20-25CODEN: MSCEEE; ISSN:0928-4931. (Elsevier B.V.)
The aim of this study was to develop chitosan composite scaffolds with high strength and controlled pore structures by homogeneously dispersed nano-sized hydroxyapatite (nano-HAp) powders. In the fabrication of composite scaffolds, nano-HAp powders distributed in an alginate (AG) soln. with a pH higher than 10 were mixed with a chitosan (CS) soln. and then freeze dried. While the HAp content increased up to 70 wt.%, the compressive strength and the elastic modulus of the composite scaffolds significantly increased from 0.27 MPa and 4.42 MPa to 0.68 MPa and 13.35 MPa, resp. Higher content of the HAp also helped develop more differentiation and mineralization of the MC3T3-E1 cells on the composite scaffolds. The uniform pore structure and the excellent mech. properties of the HAp/CS composite scaffolds likely resulted from the use of the AG soln. at pH 10 as a dispersant for the nano-HAp powders.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2MXot1WhtrY%253D&md5=ead05f536107d9eb8a0eee196dc0965c
42
Liu, Q.; Li, Q.; Xu, S.; Zheng, Q.; Cao, X. Preparation and Properties of 3D Printed Alginate–Chitosan Polyion Complex Hydrogels for Tissue Engineering. Polymers 2018, 10, 664 DOI: 10.3390/polym10060664
Google Scholar
There is no corresponding record for this reference.
43
Yu, C.-C.; Chang, J.-J.; Lee, Y.-H.; Lin, Y.-C.; Wu, M.-H.; Yang, M.-C.; Chien, C.-T. Electrospun scaffolds composing of alginate, chitosan, collagen and hydroxyapatite for applying in bone tissue engineering. Mater. Lett. 2013, 93, 133– 136, DOI: 10.1016/j.matlet.2012.11.040
Google Scholar
43
Electrospun scaffolds composing of alginate, chitosan, collagen and hydroxyapatite for applying in bone tissue engineering
Yu, Chia-Cherng; Chang, Jung-Jhih; Lee, Yen-Hsien; Lin, Yu-Cheng; Wu, Meng-Hsiu; Yang, Ming-Chien; Chien, Chiang-Ting
Materials Letters (2013), 93 (), 133-136CODEN: MLETDJ; ISSN:0167-577X. (Elsevier B.V.)
In this study, a composite scaffold made of alginate (AL), chitosan (ChS), collagen (Col), and hydroxyapatite (HAp) was fabricated by electrospinning techniques. The distribution of each component of composite scaffold was revealed by confocal laser scanning microscope (CLSM) using fluorescent labeling polymers. The morphol. and microstructure of the scaffold was examd. using a field-emission scanning electron microscope (FE-SEM) and transmission electron microscopy (TEM). To mimic the stability of these scaffolds in physiol. fluids, the degree of disintegration of collagen from these scaffolds in collagenase soln. was also tested. The results showed that the composite scaffold can greatly reduce the disintegration by 35% for 10 days in collagenase soln. Therefore, this composite is expected to be a potential scaffold for bone tissue engineering applications.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC3sXhtlahsLw%253D&md5=c76881192beef7c64389fc5b66bd43fd
44
Kolanthai, E.; Sindu, P. A.; Khajuria, D. K.; Veerla, S. C.; Kuppuswamy, D.; Catalani, L. H.; Mahapatra, D. R. Graphene Oxide─A Tool for the Preparation of Chemically Crosslinking Free Alginate–Chitosan–Collagen Scaffolds for Bone Tissue Engineering. ACS Appl. Mater. Interfaces 2018, 10 (15), 12441– 12452, DOI: 10.1021/acsami.8b00699
Google Scholar
44
Graphene oxide-A tool for the preparation of chemically crosslinking free alginate-chitosan-collagen scaffolds for bone tissue engineering
Kolanthai, Elayaraja; Sindu, Pugazhendhi Abinaya; Khajuria, Deepak Kumar; Veerla, Sarath Chandra; Kuppuswamy, Dhandapani; Catalani, Luiz Henrique; Mahapatra, D. Roy
ACS Applied Materials & Interfaces (2018), 10 (15), 12441-12452CODEN: AAMICK; ISSN:1944-8244. (American Chemical Society)
Developing a biodegradable scaffold remains a major challenge in bone tissue engineering. This study was aimed at developing novel alginate-chitosan-collagen (SA-CS-Col)-based composite scaffolds consisting of graphene oxide (GO) to enrich porous structures, elicited by the freeze-drying technique. To characterize porosity, water absorption, and compressive modulus, GO scaffolds (SA-CS-Col-GO) were prepd. with and without Ca2+-mediated crosslinking (chem. crosslinking) and analyzed using Raman, Fourier transform IR (FTIR), X-ray diffraction (XRD), and SEM techniques. The incorporation of GO into the SA-CS-Col matrix increased both crosslinking d. as indicated by the redn. of cryst. peaks in the XRD patterns and polyelectrolyte ion complex as confirmed by FTIR. GO scaffolds showed increased mech. properties which were further increased for chem. crosslinked scaffolds. All scaffolds exhibited interconnected pores of 10-250 μm range. By increasing the crosslinking d. with Ca2+, a decrease in the porosity/swelling ratio was obsd. Moreover, the SA-CS-Col-GO scaffold with or without chem. crosslinking was more stable as compared to SA-CS or SA-CS-Col scaffolds when placed in aq. soln. To perform in vitro biochem. studies, mouse osteoblast cells were grown on various scaffolds and evaluated for cell proliferation by using MTT assay and mineralization and differentiation by alizarin red S staining. These measurements showed a significant increase for cells attached to the SA-CS-Col-GO scaffold compared to SA-CS or SA-CS-Col composites. However, chem. crosslinking of SA-CS-Col-GO showed no effect on the osteogenic ability of osteoblasts. These studies indicate the potential use of GO to prep. free SA-CS-Col scaffolds with preserved porous structure with elongated Col fibrils and that these composites, which are biocompatible and stable in a biol. medium, could be used for application in engineering bone tissues.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC1cXmt1Wmu7o%253D&md5=fba4604c1893657096bb29ef1fee8d06
45
Lee, K. Y.; Mooney, D. J. Alginate: Properties and biomedical applications. Prog. Polym. Sci. 2012, 37 (1), 106– 126, DOI: 10.1016/j.progpolymsci.2011.06.003
Google Scholar
45
Alginate: Properties and biomedical applications
Lee, Kuen Yong; Mooney, David J.
Progress in Polymer Science (2012), 37 (1), 106-126CODEN: PRPSB8; ISSN:0079-6700. (Elsevier Ltd.)
A review. Alginate is a biomaterial that has found numerous applications in biomedical science and engineering due to its favorable properties, including biocompatibility and ease of gelation. Alginate hydrogels were particularly attractive in wound healing, drug delivery, and tissue engineering applications to date, as these gels retain structural similarity to the extracellular matrixes in tissues and can be manipulated to play several crit. roles. This review will provide a comprehensive overview of general properties of alginate and its hydrogels, their biomedical applications, and suggest new perspectives for future studies with these polymers.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC3MXhsVaqt77L&md5=bfd5e6059760ff18f4ee316cc28fdb8b
46
Pellá, M. C.; Lima-Tenório, M. K.; Tenório-Neto, E. T.; Guilherme, M. R.; Muniz, E. C.; Rubira, A. F. Chitosan-based hydrogels: From preparation to biomedical applications. Carbohydr. Polym. 2018, 196, 233– 245, DOI: 10.1016/j.carbpol.2018.05.033
Google Scholar
46
Chitosan-based hydrogels: From preparation to biomedical applications
Pella, Michelly C. G.; Lima-Tenorio, Michele K.; Tenorio-Neto, Ernandes T.; Guilherme, Marcos R.; Muniz, Edvani C.; Rubira, Adley F.
Carbohydrate Polymers (2018), 196 (), 233-245CODEN: CAPOD8; ISSN:0144-8617. (Elsevier Ltd.)
A review. The advances in the field of biomaterials have led to several studies on alternative biocompatible devices and to their development focusing on their properties, benefits, limitations, and utilization of alternative resources. Due to their advantages like biocompatibility, biodegradability, and low cost, polysaccharides have been widely used in the development of hydrogels. Among the polysaccharides studied on hydrogels prepn., chitosan (pure or combined with natural/synthetic polymers) have been widely investigated for use in biomedical field. In view of potential applications of chitosan-based hydrogels, this review focuses on the most recent progress made with respect to prepn., properties, and their salient accomplishments for drug delivery and tissue engineering.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC1cXpvVKgtLk%253D&md5=4f0083dd48a76550cff3518e68a86104
47
Callahan, L. A. S. Combinatorial Method/High Throughput Strategies for Hydrogel Optimization in Tissue Engineering Applications. Gels 2016, 2 (2), 18 DOI: 10.3390/gels2020018
Google Scholar
There is no corresponding record for this reference.
48
Day, E. C.; Chittari, S. S.; Bogen, M. P.; Knight, A. S. Navigating the Expansive Landscapes of Soft Materials: A User Guide for High-Throughput Workflows. ACS Polym. Au 2023, 3 (6), 406– 427, DOI: 10.1021/acspolymersau.3c00025
Google Scholar
48
Navigating the Expansive Landscapes of Soft Materials: A User Guide for High-throughput Workflows
Day, Erin C.; Chittari, Supraja S.; Bogen, Matthew P.; Knight, Abigail S.
ACS Polymers Au (2023), 3 (6), 406-427CODEN: APACCD; ISSN:2694-2453. (American Chemical Society)
Synthetic polymers are highly customizable with tailored structures and functionality, yet this versatility generates challenges in the design of advanced materials due to the size and complexity of the design space. Thus, exploration and optimization of polymer properties using combinatorial libraries has become increasingly common, which requires careful selection of synthetic strategies, characterization techniques, and rapid processing workflows to obtain fundamental principles from these large data sets. Herein, we provide guidelines for strategic design of macromol. libraries and workflows to efficiently navigate these high-dimensional design spaces. We describe synthetic methods for multiple library sizes and structures as well as characterization methods to rapidly generate data sets, including tools that can be adapted from biol. workflows. We further highlight relevant insights from statistics and machine learning to aid in data featurization, representation, and anal. This perspective acts as a "user guide" for researchers interested in leveraging high-throughput screening toward the design of multifunctional polymers and predictive modeling of structure-property relationships in soft materials.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3sXisFWgtrbF&md5=13654acd9ef4a5920a79b0f820a80a87
49
Yuan, X.; Liu, R.; Zhang, W.; Song, X.; Xu, L.; Zhao, Y.; Shang, L.; Zhang, J. Preparation of carboxylmethylchitosan and alginate blend membrane for diffusion-controlled release of diclofenac diethylamine. J. Mater. Sci. Technol. 2021, 63, 210– 215, DOI: 10.1016/j.jmst.2020.05.008
Google Scholar
49
Preparation of carboxylmethylchitosan and alginate blend membrane for diffusion-controlled release of diclofenac diethylamine
Yuan, Xiaoxue; Liu, Ran; Zhang, Wenchang; Song, Xiaoqiang; Xu, Lei; Zhao, Yan; Shang, Lei; Zhang, Jingsong
Journal of Materials Science & Technology (Shenyang, China) (2021), 63 (), 210-215CODEN: JSCTEQ; ISSN:1005-0302. (Editorial Board of Journal of Materials Science & Technology)
Controlled drug release technol. is becoming a com. sound methodol. of administering pharmaceutical therapies, and it is important to predict and control the release kinetics to fully take advantage of this technol. Carboxylmethylchitosan (CMCS) and alginate (SA) blend membranes were successfully prepd. by soln. casting technique and exposing to acetic acid atm., which induced the protonation of the amino groups of CMCS, and the formation of polyelectrolyte composite hydrogels. By using diclofenac diethylamine as the model drug, the diffusion controlled drug release behaviors of the membranes were studied based on permeation expt., which indicated that the swelling rates of the membranes could be adjusted by simply changing the wt. ratio of CMCS/SA and effectively changed the release rate of the membrane. The study shows that CMCS/SA blend membranes are promising biomaterials used for modulating the permeation behavior and developing advanced drug delivery systems.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB38XhtFehsLzM&md5=6b7821ff6a3ccf5295435cfeb5580f03
50
Wu, P.; Fang, Y.; Chen, K.; Wu, M.; Zhang, W.; Wang, S.; Liu, D.; Gao, J.; Li, H.; Lv, J.; Zhao, Y. Study of double network hydrogels based on sodium methacrylate alginate and carboxymethyl chitosan. Eur. Polym. J. 2023, 194, 112137 DOI: 10.1016/j.eurpolymj.2023.112137
Google Scholar
There is no corresponding record for this reference.
51
Urayama, K.; Takigawa, T.; Masuda, T. Poisson’s ratio of poly(vinyl alcohol) gels. Macromolecules 1993, 26 (12), 3092– 3096, DOI: 10.1021/ma00064a016
Google Scholar
There is no corresponding record for this reference.
52
Gu, Y.; Zhao, J.; Johnson, J. A. Polymer Networks. In Macromolecular Engineering: From Precise Synthesis to Macroscopic Materials and Applications; Wiley-VCH, 2022; pp 1– 52.
Google Scholar
There is no corresponding record for this reference.

Cited By

Click to copy section linkSection link copied!

This article has not yet been cited by other publications.

Download PDF

Get e-Alerts

ACS Applied Materials & Interfaces

Cite this: ACS Appl. Mater. Interfaces 2024, 16, 51, 70310–70321

Click to copy citationCitation copied!

https://doi.org/10.1021/acsami.4c16614

Published December 11, 2024

CC-BY 4.0 .

Article Views

Altmetric

Citations

Learn about these metrics

Article Views are the COUNTER-compliant sum of full text article downloads since November 2008 (both PDF and HTML) across all institutions and individuals. These metrics are regularly updated to reflect usage leading up to the last few days.

Citations are the number of other articles citing this article, calculated by Crossref and updated daily. Find more information about Crossref citation counts.

The Altmetric Attention Score is a quantitative measure of the attention that a research article has received online. Clicking on the donut icon will load a page at altmetric.com with additional details about the score and the social media presence for the given article. Find more information on the Altmetric Attention Score and how the score is calculated.

Abstract
High Resolution Image
Download MS PowerPoint Slide
Figure 1
Figure 1. (A) Illustration of the scalable accelerated materials discovery (AMD) workflow driven by autonomous experimentation (AE) and collaborative learning. A novel collaborative learning model guides the selection of formulations to be tested by collaborating clients using labeled data generated by high-throughput experimentation (HTE). (B) Flowchart for scalable AMD via AE and collaborative learning based on consensus BO using a network of clients. (C) Illustration of the difference between traditional non-collaborative BO (i.e., independent learning) and collaborative BO with consensus (i.e., collaborative learning). Created in BioRender. Liu, Y. (2024) https://BioRender.com/r26y741.
High Resolution Image
Download MS PowerPoint Slide
Figure 2
Figure 2. (A) Reaction schemes for synthesis of multi-reducing end alginate (M-alginate) and carboxymethyl chitosan sodium salt (CMCS). (B) Illustration and photographs of acid-catalyzed composite polysaccharide hydrogel cross-linking (gelation) in the presence of acetic acid (AcOH) (photographs: left = sol; right = gel; sample composition: M-alginate 0.37 wt %, CMCS 0.37 wt %, AcOH 1.72 wt %, DIW 97.54 wt %).
High Resolution Image
Download MS PowerPoint Slide
Figure 3
Figure 3. (A) Cantilever rheometer impedance spectra: (1) before measurement (air), (2) upon submersion in an M-alginate-CMCS precursor solution (sol), and (3) after gelation by reaction with AcOH for 12 h (gel). (B) Results of the HTE study that established the constraints on AcOH concentration used for M-alginate-CMCS composite hydrogel cross-linking. (C) Heat map showing the amount of AcOH dispensed in each sample of an initial 48-sample M-alginate-CMCS composite hydrogel library. (D) Real-time phase angle at resonance (φ(t)) response (i.e., raw HTC data) for the library described in (C). (E) Heat map of the steady-state φ response for the library described in (C) (12 samples were removed (i.e., row D) due to observed phase separation; thus, row D in the heat map does not contain sample information). (F) The composition–property relation for the M-alginate-CMCS composite hydrogel associated with the data in (E) presented as a ternary diagram (the DIW content of each sample is provided as Supporting Information).
High Resolution Image
Download MS PowerPoint Slide
Figure 4
Figure 4. Absolute phase angle at resonance (|φ|) (A, B) and G′ (C, D) for scalable AMD driven by AE and collaborative learning with consensus BO vs traditional active learning by BO (i.e., independent learning). Labeled data acquired using brute-force HTE were used to inform the selection of formulations for iteration 0. (E) The mean G′ from iteration 6 to 8 and the optimal G′ obtained from collaborative learning and independent learning (* indicates p < 0.05). (F) Composition–property (G′) relation for the composite hydrogel generated by collaborative vs independent learning in terms of a ternary diagram (the DIW content of each sample is provided as Supporting Information).
High Resolution Image
Download MS PowerPoint Slide
References
This article references 52 other publications.
1. 1
  Green, M. L.; Choi, C. L.; Hattrick-Simpers, J. R.; Joshi, A. M.; Takeuchi, I.; Barron, S. C.; Campo, E.; Chiang, T.; Empedocles, S.; Gregoire, J. M. Fulfilling The Promise of The Materials Genome Initiative With High-Throughput Experimental Methodologies. Appl. Phys. Rev. 2017, 4 (1), 011105 DOI: 10.1063/1.4977487
  1
  Fulfilling the promise of the materials genome initiative with high-throughput experimental methodologies
  Green, M. L.; Choi, C. L.; Hattrick-Simpers, J. R.; Joshi, A. M.; Takeuchi, I.; Barron, S. C.; Campo, E.; Chiang, T.; Empedocles, S.; Gregoire, J. M.; Kusne, A. G.; Martin, J.; Mehta, A.; Persson, K.; Trautt, Z.; Van Duren, J.; Zakutayev, A.
  Applied Physics Reviews (2017), 4 (1), 011105/1-011105/18CODEN: APRPG5; ISSN:1931-9401. (American Institute of Physics)
  The Materials Genome Initiative, a national effort to introduce new materials into the market faster and at lower cost, has made significant progress in computational simulation and modeling of materials. To build on this progress, a large amt. of exptl. data for validating these models, and informing more sophisticated ones, will be required. High-throughput experimentation generates large vols. of exptl. data using combinatorial materials synthesis and rapid measurement techniques, making it an ideal exptl. complement to bring the Materials Genome Initiative vision to fruition. This paper reviews the state-of-the-art results, opportunities, and challenges in high-throughput experimentation for materials design. A major conclusion is that an effort to deploy a federated network of high-throughput exptl. (synthesis and characterization) tools, which are integrated with a modern materials data infrastructure, is needed. (c) 2017 American Institute of Physics.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2sXltlKkurY%253D&md5=8b20fd6d59f2ef776b47e8e8a4503651
2. 2
  de Pablo, J. J.; Jackson, N. E.; Webb, M. A.; Chen, L.-Q.; Moore, J. E.; Morgan, D.; Jacobs, R.; Pollock, T.; Schlom, D. G.; Toberer, E. S. New frontiers for the materials genome initiative. npj Comput. Mater. 2019, 5 (1), 41 DOI: 10.1038/s41524-019-0173-4
  There is no corresponding record for this reference.
3. 3
  Maier, W. F.; Stöwe, K.; Sieg, S. Combinatorial and High-Throughput Materials Science. Angew. Chem., Int. Ed. 2007, 46 (32), 6016– 6067, DOI: 10.1002/anie.200603675
  3
  Combinatorial and high-throughput materials science
  Maier, Wilhelm F.; Stoewe, Klaus; Sieg, Simone
  Angewandte Chemie, International Edition (2007), 46 (32), 6016-6067, S6016/1-S6016/4CODEN: ACIEF5; ISSN:1433-7851. (Wiley-VCH Verlag GmbH & Co. KGaA)
  A review. High-through-put techniques are already used routinely for research on complex materials, such as polymers, electronic materials, and catalysts. Remarkable achievements are related to parallel syntheses and analyses, data-mining technologies, modeling approaches, as well as evolutionary strategies for materials development and formulations. The discoveries and success stories document the power of these emerging technologies.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BD2sXpsFyit7s%253D&md5=bd0bde3ae47b664b7a15120522d09b05
4. 4
  Eyke, N. S.; Koscher, B. A.; Jensen, K. F. Toward Machine Learning-Enhanced High-Throughput Experimentation. Trends Chem. 2021, 3 (2), 120– 132, DOI: 10.1016/j.trechm.2020.12.001
  4
  Toward Machine Learning-Enhanced High-Throughput Experimentation
  Eyke, Natalie S.; Koscher, Brent A.; Jensen, Klavs F.
  Trends in Chemistry (2021), 3 (2), 120-132CODEN: TCRHBQ; ISSN:2589-5974. (Cell Press)
  A review. Recent literature suggests that the fields of machine learning (ML) and high-throughput experimentation (HTE) have sep. received considerable attention from chemists and engineers, leading to the development of powerful reactivity models and platforms capable of rapidly performing thousands of reactions. The merger of ML with HTE presents a wealth of opportunities for the exploration of chem. space, but the integration of the two has yet to be fully realized. We highlight examples of recent developments in ML and HTE that collectively suggest the utility of their integration. Our anal. highlights the complementarity of the two fields, while exposing a no. of obstacles that can and should be overcome to take full advantage of this merger and thereby accelerate chem. research.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3MXhtFeltrnO&md5=231220ca97d31091f9d83a7a246fcb27
5. 5
  Häse, F.; Roch, L. M.; Aspuru-Guzik, A. Next-Generation Experimentation with Self-Driving Laboratories. Trends Chem. 2019, 1 (3), 282– 291, DOI: 10.1016/j.trechm.2019.02.007
  5
  Next-Generation Experimentation with Self-Driving Laboratories
  Hase, Florian; Roch, Loic M.; Aspuru-Guzik, Alan
  Trends in Chemistry (2019), 1 (3), 282-291CODEN: TCRHBQ; ISSN:2589-5974. (Cell Press)
  A review. The ever-growing demand for advanced functional materials requires disruption of conventional approaches to experimentation and acceleration of the discovery process. State-of-the-art approaches to scientific discovery are inherently slow, capital intensive, and have arguably reached a plateau. Significant advances are possible when rethinking and redesigning the traditional experimentation process. Self-driving labs. promise to substantially accelerate the discovery process by augmenting automated experimentation platforms with artificial intelligence (AI). AI methods actively search for promising exptl. procedures by hypothesizing about their outcomes based on previous expts. This feedback loop is crucial to reduce the no. of expts. needed for discovery. Supplying automated platforms with AI enables self-driving labs. to fully embrace the vision of autonomous experimentation.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3cXitFehtL3L&md5=8118dcb8ae47482044b08d47d0f43952
6. 6
  Abolhasani, M.; Kumacheva, E. The rise of self-driving labs in chemical and materials sciences. Nat. Synth. 2023, 2 (6), 483– 492, DOI: 10.1038/s44160-022-00231-0
  There is no corresponding record for this reference.
7. 7
  Liu, Y.; Zhang, J.; Zhang, Y.; Yoon, H. Y.; Jia, X.; Roman, M.; Johnson, B. N. Accelerated Engineering of Optimized Functional Composite Hydrogels via High-Throughput Experimentation. ACS Appl. Mater. Interfaces 2023, 15 (45), 52908– 52920, DOI: 10.1021/acsami.3c11483
  There is no corresponding record for this reference.
8. 8
  Park, T.; Kim, E.; Sun, J.; Kim, M.; Hong, E.; Min, K. Rapid discovery of promising materials via active learning with multi-objective optimization. Mater. Today Commun. 2023, 37, 107245 DOI: 10.1016/j.mtcomm.2023.107245
  There is no corresponding record for this reference.
9. 9
  Bai, Y.; Khoo, Z. H. J.; Made, R. I.; Xie, H.; Lim, C. Y. J.; Handoko, A. D.; Chellappan, V.; Cheng, J. J.; Wei, F.; Lim, Y.-F. Closed Loop Multi-Objective Optimization for Cu-Sb-S Photoelectrocatalytic Materials Discovery. Adv. Mater. 2023, 36 (2), 2304269 DOI: 10.1002/adma.202304269
  There is no corresponding record for this reference.
10. 10
  Orlova, T.; Piven, A.; Darmoroz, D.; Aliev, T.; Razik, T.; Boitsev, A.; Grafeeva, N.; Skorb, E. Machine learning for soft and liquid molecular materials. Digital Discovery 2023, 2 (2), 298– 315, DOI: 10.1039/D2DD00132B
  There is no corresponding record for this reference.
11. 11
  Li, Z. H.; Song, P. R.; Li, G. F.; Han, Y. F.; Ren, X. X.; Bai, L.; Su, J. C. AI energized hydrogel design, optimization and application in biomedicine. Mater. Today Bio 2024, 25, 101014 DOI: 10.1016/j.mtbio.2024.101014
  There is no corresponding record for this reference.
12. 12
  Oliveira, M. B.; Mano, J. F. High-throughput screening for integrative biomaterials design: exploring advances and new trends. Trends Biotechnol. 2014, 32 (12), 627– 636, DOI: 10.1016/j.tibtech.2014.09.009
  There is no corresponding record for this reference.
13. 13
  Callahan, L. S. Combinatorial Method/High Throughput Strategies for Hydrogel Optimization in Tissue Engineering Applications. Gels 2016, 2 (2), 18 DOI: 10.3390/gels2020018
  There is no corresponding record for this reference.
14. 14
  Green, M. L.; Choi, C. L.; Hattrick-Simpers, J. R.; Joshi, A. M.; Takeuchi, I.; Barron, S. C.; Campo, E.; Chiang, T.; Empedocles, S.; Gregoire, J. M. Fulfilling the promise of the materials genome initiative with high-throughput experimental methodologies. Appl. Phys. Rev. 2017, 4 (1), 011105 DOI: 10.1063/1.4977487
  14
  Fulfilling the promise of the materials genome initiative with high-throughput experimental methodologies
  Green, M. L.; Choi, C. L.; Hattrick-Simpers, J. R.; Joshi, A. M.; Takeuchi, I.; Barron, S. C.; Campo, E.; Chiang, T.; Empedocles, S.; Gregoire, J. M.; Kusne, A. G.; Martin, J.; Mehta, A.; Persson, K.; Trautt, Z.; Van Duren, J.; Zakutayev, A.
  Applied Physics Reviews (2017), 4 (1), 011105/1-011105/18CODEN: APRPG5; ISSN:1931-9401. (American Institute of Physics)
  The Materials Genome Initiative, a national effort to introduce new materials into the market faster and at lower cost, has made significant progress in computational simulation and modeling of materials. To build on this progress, a large amt. of exptl. data for validating these models, and informing more sophisticated ones, will be required. High-throughput experimentation generates large vols. of exptl. data using combinatorial materials synthesis and rapid measurement techniques, making it an ideal exptl. complement to bring the Materials Genome Initiative vision to fruition. This paper reviews the state-of-the-art results, opportunities, and challenges in high-throughput experimentation for materials design. A major conclusion is that an effort to deploy a federated network of high-throughput exptl. (synthesis and characterization) tools, which are integrated with a modern materials data infrastructure, is needed. (c) 2017 American Institute of Physics.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2sXltlKkurY%253D&md5=8b20fd6d59f2ef776b47e8e8a4503651
15. 15
  Di Fiore, F.; Nardelli, M.; Mainini, L. Active Learning and Bayesian Optimization: A Unified Perspective to Learn with a Goal. Arch. Comput. Methods Eng. 2024, 31, 2985– 3013, DOI: 10.1007/s11831-024-10064-z
  There is no corresponding record for this reference.
16. 16
  Settles, B. Active Learning; Morgan & Claypool Publishers, 2012.
  There is no corresponding record for this reference.
17. 17
  Kusne, A. G.; Yu, H.; Wu, C.; Zhang, H.; Hattrick-Simpers, J.; DeCost, B.; Sarker, S.; Oses, C.; Toher, C.; Curtarolo, S. On-the-fly closed-loop materials discovery via Bayesian active learning. Nat. Commun. 2020, 11 (1), 5966 DOI: 10.1038/s41467-020-19597-w
  17
  On-the-fly closed-loop materials discovery via Bayesian active learning
  Kusne, A. Gilad; Yu, Heshan; Wu, Changming; Zhang, Huairuo; Hattrick-Simpers, Jason; DeCost, Brian; Sarker, Suchismita; Oses, Corey; Toher, Cormac; Curtarolo, Stefano; Davydov, Albert V.; Agarwal, Ritesh; Bendersky, Leonid A.; Li, Mo; Mehta, Apurva; Takeuchi, Ichiro
  Nature Communications (2020), 11 (1), 5966CODEN: NCAOBW; ISSN:2041-1723. (Nature Research)
  Active learning-the field of machine learning (ML) dedicated to optimal expt. design-has played a part in science as far back as the 18th century when Laplace used it to guide his discovery of celestial mechanics. In this work, we focus a closed-loop, active learning-driven autonomous system on another major challenge, the discovery of advanced materials against the exceedingly complex synthesis-processes-structure-property landscape. We demonstrate an autonomous materials discovery methodol. for functional inorg. compds. which allow scientists to fail smarter, learn faster, and spend less resources in their studies, while simultaneously improving trust in scientific results and machine learning tools. This robot science enables science-over-the-network, reducing the economic impact of scientists being phys. sepd. from their labs. The real-time closed-loop, autonomous system for materials exploration and optimization (CAMEO) is implemented at the synchrotron beamline to accelerate the interconnected tasks of phase mapping and property optimization, with each cycle taking seconds to minutes. We also demonstrate an embodiment of human-machine interaction, where human-in-the-loop is called to play a contributing role within each cycle. This work has resulted in the discovery of a novel epitaxial nanocomposite phase-change memory material.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3cXisVOnsL3N&md5=138fb31700fa99c1962c048fe7829145
18. 18
  Min, K.; Cho, E. Accelerated Discovery of Novel Inorganic Materials with Desired Properties Using Active Learning. J. Phys. Chem. C 2020, 124 (27), 14759– 14767, DOI: 10.1021/acs.jpcc.0c00545
  18
  Accelerated Discovery of Novel Inorganic Materials with Desired Properties Using Active Learning
  Min, Kyoungmin; Cho, Eunseog
  Journal of Physical Chemistry C (2020), 124 (27), 14759-14767CODEN: JPCCCK; ISSN:1932-7447. (American Chemical Society)
  Construction of prediction models using machine learning algorithms on existing databases expands the search limit of undiscovered structures, in principle, to the entire materials space. However, because of uncertainties in machine learning prediction, the suggested properties are not always promising; thus, improving the database quality is mandatory for validation as well as improvement in prediction accuracy. To achieve this, we herein implement an active learning process, beginning with a limited no. of databases, to find materials satisfying target properties (band gap and refractive index) with minimized trials and errors. The regression model is initially trained with only around 2% of the entire search space, and 20 new databases, suggested from the optimization schemes, are added at each optimization process. Between exploration, exploitation, random selection, and the Bayesian optimization method, the Bayesian method exhibits the best performance in finding the no. of materials that satisfies the criteria within limited trials In addn., the structure with the max. target property values is found after searching only around 7.0% and 7.7% of the entire database for band gap and refractive index, resp. Current results clearly confirm that the active learning process can be accelerated to find ideal materials satisfying target properties with minimized resources.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3cXht1Git77J&md5=92537da74ec4d8a89d711934b7a91114
19. 19
  Oftelie, L. B.; Rajak, P.; Kalia, R. K.; Nakano, A.; Sha, F.; Sun, J.; Singh, D. J.; Aykol, M.; Huck, P.; Persson, K.; Vashishta, P. Active learning for accelerated design of layered materials. npj Comput. Mater. 2018, 4 (1), 74 DOI: 10.1038/s41524-018-0129-0
  There is no corresponding record for this reference.
20. 20
  Shahriari, B.; Swersky, K.; Wang, Z.; Adams, R. P.; Freitas, N. d. Taking the Human Out of the Loop: A Review of Bayesian Optimization. Proc. IEEE 2016, 104 (1), 148– 175, DOI: 10.1109/jproc.2015.2494218
  There is no corresponding record for this reference.
21. 21
  Chen, H.; Zheng, L.; Kontar, R. A.; Raskutti, G. Gaussian process parameter estimation using mini-batch stochastic gradient descent: convergence guarantees and empirical benefits. J. Mach. Learn. Res. 2022, 23 (1), 1– 59
  There is no corresponding record for this reference.
22. 22
  Rasmussen, C. E.; Williams, C. K. I. Gaussian Processes for Machine Learning; The MIT Press, 2005.
  There is no corresponding record for this reference.
23. 23
  Burger, B.; Maffettone, P. M.; Gusev, V. V.; Aitchison, C. M.; Bai, Y.; Wang, X.; Li, X.; Alston, B. M.; Li, B.; Clowes, R. A mobile robotic chemist. Nature 2020, 583 (7815), 237– 241, DOI: 10.1038/s41586-020-2442-2
  23
  A mobile robotic chemist
  Burger, Benjamin; Maffettone, Phillip M.; Gusev, Vladimir V.; Aitchison, Catherine M.; Bai, Yang; Wang, Xiaoyan; Li, Xiaobo; Alston, Ben M.; Li, Buyi; Clowes, Rob; Rankin, Nicola; Harris, Brandon; Sprick, Reiner Sebastian; Cooper, Andrew I.
  Nature (London, United Kingdom) (2020), 583 (7815), 237-241CODEN: NATUAS; ISSN:0028-0836. (Nature Research)
  Abstr.: Technologies such as batteries, biomaterials and heterogeneous catalysts have functions that are defined by mixts. of mol. and mesoscale components. As yet, this multi-length-scale complexity cannot be fully captured by atomistic simulations, and the design of such materials from first principles is still rare1-5. Likewise, exptl. complexity scales exponentially with the no. of variables, restricting most searches to narrow areas of materials space. Robots can assist in exptl. searches6-14 but their widespread adoption in materials research is challenging because of the diversity of sample types, operations, instruments and measurements required. Here we use a mobile robot to search for improved photocatalysts for hydrogen prodn. from water15. The robot operated autonomously over eight days, performing 688 expts. within a ten-variable exptl. space, driven by a batched Bayesian search algorithm16-18. This autonomous search identified photocatalyst mixts. that were six times more active than the initial formulations, selecting beneficial components and deselecting neg. ones. Our strategy uses a dexterous19,20 free-roaming robot21-24, automating the researcher rather than the instruments. This modular approach could be deployed in conventional labs. for a range of research problems beyond photocatalysis.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3cXhtlCqur%252FN&md5=6a13f9399b94cca78cdeac312a2e4563
24. 24
  Shields, B. J.; Stevens, J.; Li, J.; Parasram, M.; Damani, F.; Alvarado, J. I. M.; Janey, J. M.; Adams, R. P.; Doyle, A. G. Bayesian reaction optimization as a tool for chemical synthesis. Nature 2021, 590 (7844), 89– 96, DOI: 10.1038/s41586-021-03213-y
  24
  Bayesian reaction optimization as a tool for chemical synthesis
  Shields, Benjamin J.; Stevens, Jason; Li, Jun; Parasram, Marvin; Damani, Farhan; Alvarado, Jesus I. Martinez; Janey, Jacob M.; Adams, Ryan P.; Doyle, Abigail G.
  Nature (London, United Kingdom) (2021), 590 (7844), 89-96CODEN: NATUAS; ISSN:0028-0836. (Nature Research)
  Abstr.: Reaction optimization is fundamental to synthetic chem., from optimizing the yield of industrial processes to selecting conditions for the prepn. of medicinal candidates1. Likewise, parameter optimization is omnipresent in artificial intelligence, from tuning virtual personal assistants to training social media and product recommendation systems2. Owing to the high cost assocd. with carrying out expts., scientists in both areas set numerous (hyper)parameter values by evaluating only a small subset of the possible configurations. Bayesian optimization, an iterative response surface-based global optimization algorithm, has demonstrated exceptional performance in the tuning of machine learning models3. Bayesian optimization has also been recently applied in chem.4-9; however, its application and assessment for reaction optimization in synthetic chem. has not been investigated. Here we report the development of a framework for Bayesian reaction optimization and an open-source software tool that allows chemists to easily integrate state-of-the-art optimization algorithms into their everyday lab. practices. We collect a large benchmark dataset for a palladium-catalyzed direct arylation reaction, perform a systematic study of Bayesian optimization compared to human decision-making in reaction optimization, and apply Bayesian optimization to two real-world optimization efforts (Mitsunobu and deoxyfluorination reactions). Benchmarking is accomplished via an online game that links the decisions made by expert chemists and engineers to real expts. run in the lab. Our findings demonstrate that Bayesian optimization outperforms human decisionmaking in both av. optimization efficiency (no. of expts.) and consistency (variance of outcome against initially available data). Overall, our studies suggest that adopting Bayesian optimization methods into everyday lab. practices could facilitate more efficient synthesis of functional chems. by enabling better-informed, data-driven decisions about which expts. to run.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3MXjt1SjtLg%253D&md5=4ae6e5eb6ae548b86f651b4a72508e28
25. 25
  Kusne, A. G.; McDannald, A. Scalable multi-agent lab framework for lab optimization. Matter 2023, 6 (6), 1880– 1893, DOI: 10.1016/j.matt.2023.03.022
  There is no corresponding record for this reference.
26. 26
  Kontar, R.; Shi, N. C.; Yue, X. B.; Chung, S.; Byon, E.; Chowdhury, M.; Jin, J. H.; Kontar, W.; Masoud, N.; Nouiehed, M. The Internet of Federated Things (IoFT). IEEE Access 2021, 9, 156071– 156113, DOI: 10.1109/ACCESS.2021.3127448
  There is no corresponding record for this reference.
27. 27
  Frazier, P. I. A. Tutorial on Bayesian optimization, arXiv:1807.02811. arXiv.org e-Print archive, 2018. https://arXiv.org/abs/1807.02811.
  There is no corresponding record for this reference.
28. 28
  Gardner, J. R.; Pleiss, G.; Bindel, D.; Weinberger, K. Q.; Wilson, A. G. GPyTorch: Blackbox Matrix-Matrix Gaussian Process Inference with GPU Acceleration, Advances in Neural Information Processing Systems; NeurIPS, 2018.
  There is no corresponding record for this reference.
29. 29
  Yue, X.; Al Kontar, R.; Berahas, A. S.; Liu, Y.; Zai, Z.; Edgar, K.; Johnson, B. N. Collaborative and Distributed Bayesian Optimization via Consensus: Showcasing the Power of Collaboration for Optimal Design, arXiv:2306.14348. arXiv.org e-Print archive, 2023. https://arXiv.org/abs/2306.14348.
  There is no corresponding record for this reference.
30. 30
  Balandat, M.; Karrer, B.; Jiang, D. R.; Daulton, S.; Letham, B.; Wilson, A. G.; Bakshy, E. BOTORCH: A Framework for Efficient Monte-Carlo Bayesian Optimization, Advances in Neural Information Processing Systems; NeurIPS, 2020.
  There is no corresponding record for this reference.
31. 31
  Zhai, Z.; Zhou, Y.; Korovich, A. G.; Hall, B. A.; Yoon, H. Y.; Yao, Y.; Zhang, J.; Bortner, M. J.; Roman, M.; Madsen, L. A.; Edgar, K. J. Synthesis and Characterization of Multi-Reducing-End Polysaccharides. Biomacromolecules 2023, 24 (6), 2596– 2605, DOI: 10.1021/acs.biomac.3c00104
  There is no corresponding record for this reference.
32. 32
  Zhou, Y.; Zhai, Z.; Yao, Y.; Stant, J. C.; Landrum, S. L.; Bortner, M. J.; Frazier, C. E.; Edgar, K. J. Oxidized hydroxypropyl cellulose/carboxymethyl chitosan hydrogels permit pH-responsive, targeted drug release. Carbohydr. Polym. 2023, 300, 120213 DOI: 10.1016/j.carbpol.2022.120213
  32
  Oxidized hydroxypropyl cellulose/carboxymethyl chitosan hydrogels permit pH-responsive, targeted drug release
  Zhou, Yang; Zhai, Zhenghao; Yao, Yimin; Stant, John C.; Landrum, Sarah L.; Bortner, Michael J.; Frazier, Charles E.; Edgar, Kevin J.
  Carbohydrate Polymers (2023), 300 (), 120213CODEN: CAPOD8; ISSN:0144-8617. (Elsevier Ltd.)
  Polysaccharide-based Schiff base hydrogels have promise for drug delivery, tissue engineering, and many other applications due to their reversible imine bond crosslinks. We describe herein pH-responsive, injectable, and self-healing hydrogels prepd. by reacting oxidized hydroxypropyl cellulose (Ox-HPC) with carboxymethyl chitosan (CMCS). Simple combination of ketones from Ox-HPC side chains with amines from CMCS in water provides a dynamic, hydrophilic polysaccharide network. The reversible nature of these imine bonds in the presence of water provides a hydrogel with injectable and self-healing properties. Phenylalanine as a model amine-contg. drug was linked by imine bonds to Ox-HPC within the hydrogel. Phenylalanine release was faster at the pH of the extracellular space around tumors (6.8) than in normal tissues (7.4), a surprising degree of pH sensitivity. Therefore, Ox-HPC/CMCS hydrogels show promise as drug carriers that may selectively target even slightly lower pH environments like the extracellular milieu around cancer cells.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB38Xisleisb%252FP&md5=f93efc00c7e1e21a973a0925d701a6d8
33. 33
  Zhang, J.; Liu, Y.; Sekhar P, D. C.; Singh, M.; Tong, Y.; Kucukdeger, E.; Yoon, H. Y.; Haring, A. P.; Roman, M.; Kong, Z.; Johnson, B. N. Rapid, autonomous high-throughput characterization of hydrogel rheological properties via automated sensing and physics-guided machine learning. Appl. Mater. Today 2023, 30, 101720 DOI: 10.1016/j.apmt.2022.101720
  There is no corresponding record for this reference.
34. 34
  Haring, A. P.; Singh, M.; Koh, M.; Cesewski, E.; Dillard, D. A.; Kong, Z. J.; Johnson, B. N. Real-time characterization of hydrogel viscoelastic properties and sol-gel phase transitions using cantilever sensors. J. Rheol. 2020, 64 (4), 837– 850, DOI: 10.1122/8.0000009
  34
  Real-time characterization of hydrogel viscoelastic properties and sol-gel phase transitions using cantilever sensors
  Haring, Alexander P.; Singh, Manjot; Koh, Miharu; Cesewski, Ellen; Dillard, David A.; Kong, Zhenyu "James"; Johnson, Blake N.
  Journal of Rheology (Melville, NY, United States) (2020), 64 (4), 837-850CODEN: JORHD2; ISSN:0148-6055. (American Institute of Physics)
  Here, we report for the first time that resonance in dynamic-mode cantilever sensors persists in hydrogels and enables the real-time characterization of hydrogel viscoelastic properties and the continuous monitoring of sol-gel phase transitions (i.e., gelation and dissoln. processes). Real-time tracking of piezoelec.-excited millimeter cantilever (PEMC) sensor resonant frequency (fair = 55.4 ± 8.8 kHz; n = 5 sensors) and quality factor (Q; Qair = 23.8 ± 1.5) enabled continuous monitoring of high-frequency hydrogel shear storage and loss moduli (G'f and G"f, resp.) calcd. by sensor data and fluid-structure interaction models. Changes in the sensor phase angle, quality factor, and high-frequency shear moduli obtained at the resonant frequency (G'fand G"f) correlated with low-frequency moduli obtained at 1 Hz using dynamic mech. anal. Characterization studies were performed using phys. and chem. crosslinked hydrogel systems, including gelatin hydrogels (6-10 wt.%) and alginate hydrogels (0.25-0.75 wt.%). The sensor exhibited a dynamic range from the rheol. properties of inviscid solns. to hydrogels with high-frequency moduli of 80 kPa and low-frequency moduli of 26 kPa. The sensor exhibited a limit of detection of 260 Pa and 1.9 kPa for changes in hydrogel storage modulus (E') based on the sensor's phase angle and quality factor responses, resp. We also show that sensor data enable quant. characterization of gelation process dynamics using a modified Hill model. This work suggests that cantilever sensors provide a promising platform for the sensor-based characterization of hydrogels, such as quantification of viscoelastic properties and real-time monitoring of gelation processes. (c) 2020 American Institute of Physics.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3cXps1Ohurw%253D&md5=a263d7613603732f7e688f0ca9d15282
35. 35
  Singh, M.; Zhang, J.; Bethel, K.; Liu, Y.; Davis, E. M.; Zeng, H.; Kong, Z.; Johnson, B. N. Closed-Loop Controlled Photopolymerization of Hydrogels. ACS Appl. Mater. Interfaces 2021, 13 (34), 40365– 40378, DOI: 10.1021/acsami.1c11779
  35
  Closed-Loop Controlled Photopolymerization of Hydrogels
  Singh, Manjot; Zhang, Junru; Bethel, Keturah; Liu, Yang; Davis, Eric M.; Zeng, Haibo; Kong, Zhenyu; Johnson, Blake N.
  ACS Applied Materials & Interfaces (2021), 13 (34), 40365-40378CODEN: AAMICK; ISSN:1944-8244. (American Chemical Society)
  Here, we present a closed-loop controlled photopolymn. process for fabrication of hydrogels with controlled storage moduli. Hydrogel crosslinking was assocd. with a significant change in the phase angle of a piezoelec. cantilever sensor and established the timescale of the photopolymn. process. The compn., structure, and mech. properties of the fabricated hydrogels were characterized using Raman spectroscopy, SEM (SEM), and dynamic mech. anal. (DMA). We found that the storage moduli of photocured poly(ethylene glycol) dimethacrylate (PEGDMA) and poly(N-isopropylacrylamide) (PNIPAm) hydrogels could be controlled using bang-bang and fuzzy logic controllers. Bang-bang controlled photopolymn. resulted in const. overshoot of the storage modulus setpoint for PEGDMA hydrogels, which was mitigated by setpoint correction and fuzzy logic control. SEM and DMA studies showed that the network structure and storage modulus of PEGDMA hydrogels were dependent on the cure time and temporal profile of UV exposure during photopolymn. This work provides an advance in pulsed and continuous photopolymn. processes for hydrogel engineering based on closed-loop control that enables reproducible fabrication of hydrogels with controlled mech. properties.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3MXhvVart7fO&md5=1522b64d1d8f3d22093bf2c407f8ede0
36. 36
  Liu, Y.; Bethel, K.; Singh, M.; Zhang, J.; Ashkar, R.; Davis, E. M.; Johnson, B. N. Comparison of Bulk- vs Layer-by-Layer-Cured Stimuli-Responsive PNIPAM–Alginate Hydrogel Dynamic Viscoelastic Property Response via Embedded Sensors. ACS Appl. Polym. Mater. 2022, 4 (8), 5596– 5607, DOI: 10.1021/acsapm.2c00634
  There is no corresponding record for this reference.
37. 37
  Mather, M. L.; Rides, M.; Allen, C. R. G.; Tomlins, P. E. Liquid Viscoelasticity Probed by a Mesoscale Piezoelectric Bimorph Cantilever. J. Rheol. 2012, 56 (1), 99– 112, DOI: 10.1122/1.3670732
  37
  Liquid viscoelasticity probed by a mesoscale piezoelectric bimorph cantilever
  Mather, Melissa L.; Rides, Martin; Allen, Crispin R. G.; Tomlins, Paul E.
  Journal of Rheology (Melville, NY, United States) (2012), 56 (1), 99-112CODEN: JORHD2; ISSN:0148-6055. (American Institute of Physics)
  The viscoelastic properties of fluids are key to their performance in industries ranging from biotechnol. to the automotive industry. Traditionally, fluid viscoelastic properties are monitored with rheometers but these are expensive, require a skilled operator, function over a relatively limited frequency range and are not suitable for in situ monitoring. Piezoelec. cantilevers capable of in situ assessment of the rheol. properties of relatively small fluid vols. have the potential to overcome many of these limitations and can be fabricated into low cost probes. Rheol. assessment of test fluids using piezoelec. cantilevers is typically made through anal. of the cantilever's resonant oscillation in the fluids. For accurate results, the damping of the cantilever should be low as quantified by a high quality factor Q. This can be difficult in fluids of high viscosity particularly for microscopic cantilevers. In this paper, a "mesoscale" piezoelec. bimorph cantilever was used. The mesoscale refers to a size regime intermediate between microscopic and macroscopic, in this work the cantilever used has dimensions of the order of millimeters. This mesoscale cantilever displayed a sufficiently high Q to probe the rheol. properties of highly damping and elastic fluids in situ. The developed probe will be ideally suited to in-process monitoring of high value products such as those in the biotechnol. industry. (c) 2012 American Institute of Physics.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC38XntF2rtQ%253D%253D&md5=609f9c33b3c1cc15bc68d73d6e927421
38. 38
  Wei-Liem, L. On Latin hypercube sampling. Ann. Stat. 1996, 24 (5), 2058– 2080, DOI: 10.1214/aos/1069362310
  There is no corresponding record for this reference.
39. 39
  Yang, Q.; Peng, J.; Xiao, H.; Xu, X.; Qian, Z. Polysaccharide hydrogels: Functionalization, construction and served as scaffold for tissue engineering. Carbohydr. Polym. 2022, 278, 118952 DOI: 10.1016/j.carbpol.2021.118952
  39
  Polysaccharide hydrogels: Functionalization, construction and served as scaffold for tissue engineering
  Yang, Qian; Peng, Jinrong; Xiao, Haitao; Xu, Xuewen; Qian, Zhiyong
  Carbohydrate Polymers (2022), 278 (), 118952CODEN: CAPOD8; ISSN:0144-8617. (Elsevier Ltd.)
  A review. Polysaccharide hydrogels have been widely utilized in tissue engineering. They interact with the organismal environments, modulating the cargos release and realizing of long-term survival and activations of living cells. In this review, the potential strategies for modification of polysaccharides were introduced firstly. It is not only used to functionalize the polysaccharides for the consequent formation of hydrogels, but also used to introduce versatile side groups for the regulation of cell behavior. Then, techniques and underlying mechanisms in inducing the formation of hydrogels by polysaccharides or their derivs. are briefly summarized. Finally, the applications of polysaccharide hydrogels in vivo, mainly focus on the performance for alleviation of foreign-body response (FBR) and as cell scaffolds for tissue regeneration, are exemplified. In addn., the perspectives and challenges for further research are addressed. It aims to provide a comprehensive framework about the potentials and challenges that the polysaccharide hydrogels confronting in tissue engineering.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3MXislelsbvL&md5=301eb9a47feab5504afaa492b72872ea
40. 40
  Zhang, M.; Ma, H.; Wang, X.; Yu, B.; Cong, H.; Shen, Y. Polysaccharide-based nanocarriers for efficient transvascular drug delivery. J. Controlled Release 2023, 354, 167– 187, DOI: 10.1016/j.jconrel.2022.12.051
  40
  Polysaccharide-based nanocarriers for efficient transvascular drug delivery
  Zhang, Min; Ma, He; Wang, Xijie; Yu, Bing; Cong, Hailin; Shen, Youqing
  Journal of Controlled Release (2023), 354 (), 167-187CODEN: JCREEC; ISSN:0168-3659. (Elsevier B.V.)
  A review. Polysaccharide-based nanocarriers (PBNs) are the focus of extensive investigation because of their biocompatibility, low cost, wide availability, and chem. versatility, which allow a wide range of anticancer agents to be loaded within the nanocarriers. Similar to other nanocarriers, most PBNs are designed to extravasate out of tumor vessels, depending on the enhanced permeability and retention (EPR) effect. However, the EPR effect is compromised in some tumors due to the heterogeneity of tumor structures. Transvascular transport efficacy is decreased by complex blood vessels and condensed tumor stroma. The limited extravasation impedes efficient drug delivery into tumor parenchyma, and thus affects the subsequent tumor accumulation, which hinders the therapeutic effect of PBNs. Therefore, overcoming the biol. barriers that restrict extravasation from tumor vessels is of great importance in PBN design. Many strategies have been developed to enhance the EPR effect that involve nanocarrier property regulation and tumor structure remodeling. Moreover, some researchers have proposed active transcytosis pathways that are complementary to the paracellular EPR effect to increase the transvascular extravasation efficiency of PBNs. In this review, we summarize the recent advances in the design of PBNs with enhanced transvascular transport to enable optimization of PBNs in the extravasation of the drug delivery process. We also discuss the obstacles and challenges that need to be addressed to clarify the transendothemial mechanism of PBNs and the potential interactions between extravasation and other drug delivery steps.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3sXnsl2ntQ%253D%253D&md5=17215726f9de1cfb4c60de184cb94bda
41. 41
  Kim, H.-L.; Jung, G.-Y.; Yoon, J.-H.; Han, J.-S.; Park, Y.-J.; Kim, D.-G.; Zhang, M.; Kim, D.-J. Preparation and characterization of nano-sized hydroxyapatite/alginate/chitosan composite scaffolds for bone tissue engineering. Mater. Sci. Eng.: C 2015, 54, 20– 25, DOI: 10.1016/j.msec.2015.04.033
  41
  Preparation and characterization of nano-sized hydroxyapatite/alginate/chitosan composite scaffolds for bone tissue engineering
  Kim, Hye-Lee; Jung, Gil-Yong; Yoon, Jun-Ho; Han, Jung-Suk; Park, Yoon-Jeong; Kim, Do-Gyoon; Zhang, Miqin; Kim, Dae-Joon
  Materials Science & Engineering, C: Materials for Biological Applications (2015), 54 (), 20-25CODEN: MSCEEE; ISSN:0928-4931. (Elsevier B.V.)
  The aim of this study was to develop chitosan composite scaffolds with high strength and controlled pore structures by homogeneously dispersed nano-sized hydroxyapatite (nano-HAp) powders. In the fabrication of composite scaffolds, nano-HAp powders distributed in an alginate (AG) soln. with a pH higher than 10 were mixed with a chitosan (CS) soln. and then freeze dried. While the HAp content increased up to 70 wt.%, the compressive strength and the elastic modulus of the composite scaffolds significantly increased from 0.27 MPa and 4.42 MPa to 0.68 MPa and 13.35 MPa, resp. Higher content of the HAp also helped develop more differentiation and mineralization of the MC3T3-E1 cells on the composite scaffolds. The uniform pore structure and the excellent mech. properties of the HAp/CS composite scaffolds likely resulted from the use of the AG soln. at pH 10 as a dispersant for the nano-HAp powders.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2MXot1WhtrY%253D&md5=ead05f536107d9eb8a0eee196dc0965c
42. 42
  Liu, Q.; Li, Q.; Xu, S.; Zheng, Q.; Cao, X. Preparation and Properties of 3D Printed Alginate–Chitosan Polyion Complex Hydrogels for Tissue Engineering. Polymers 2018, 10, 664 DOI: 10.3390/polym10060664
  There is no corresponding record for this reference.
43. 43
  Yu, C.-C.; Chang, J.-J.; Lee, Y.-H.; Lin, Y.-C.; Wu, M.-H.; Yang, M.-C.; Chien, C.-T. Electrospun scaffolds composing of alginate, chitosan, collagen and hydroxyapatite for applying in bone tissue engineering. Mater. Lett. 2013, 93, 133– 136, DOI: 10.1016/j.matlet.2012.11.040
  43
  Electrospun scaffolds composing of alginate, chitosan, collagen and hydroxyapatite for applying in bone tissue engineering
  Yu, Chia-Cherng; Chang, Jung-Jhih; Lee, Yen-Hsien; Lin, Yu-Cheng; Wu, Meng-Hsiu; Yang, Ming-Chien; Chien, Chiang-Ting
  Materials Letters (2013), 93 (), 133-136CODEN: MLETDJ; ISSN:0167-577X. (Elsevier B.V.)
  In this study, a composite scaffold made of alginate (AL), chitosan (ChS), collagen (Col), and hydroxyapatite (HAp) was fabricated by electrospinning techniques. The distribution of each component of composite scaffold was revealed by confocal laser scanning microscope (CLSM) using fluorescent labeling polymers. The morphol. and microstructure of the scaffold was examd. using a field-emission scanning electron microscope (FE-SEM) and transmission electron microscopy (TEM). To mimic the stability of these scaffolds in physiol. fluids, the degree of disintegration of collagen from these scaffolds in collagenase soln. was also tested. The results showed that the composite scaffold can greatly reduce the disintegration by 35% for 10 days in collagenase soln. Therefore, this composite is expected to be a potential scaffold for bone tissue engineering applications.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC3sXhtlahsLw%253D&md5=c76881192beef7c64389fc5b66bd43fd
44. 44
  Kolanthai, E.; Sindu, P. A.; Khajuria, D. K.; Veerla, S. C.; Kuppuswamy, D.; Catalani, L. H.; Mahapatra, D. R. Graphene Oxide─A Tool for the Preparation of Chemically Crosslinking Free Alginate–Chitosan–Collagen Scaffolds for Bone Tissue Engineering. ACS Appl. Mater. Interfaces 2018, 10 (15), 12441– 12452, DOI: 10.1021/acsami.8b00699
  44
  Graphene oxide-A tool for the preparation of chemically crosslinking free alginate-chitosan-collagen scaffolds for bone tissue engineering
  Kolanthai, Elayaraja; Sindu, Pugazhendhi Abinaya; Khajuria, Deepak Kumar; Veerla, Sarath Chandra; Kuppuswamy, Dhandapani; Catalani, Luiz Henrique; Mahapatra, D. Roy
  ACS Applied Materials & Interfaces (2018), 10 (15), 12441-12452CODEN: AAMICK; ISSN:1944-8244. (American Chemical Society)
  Developing a biodegradable scaffold remains a major challenge in bone tissue engineering. This study was aimed at developing novel alginate-chitosan-collagen (SA-CS-Col)-based composite scaffolds consisting of graphene oxide (GO) to enrich porous structures, elicited by the freeze-drying technique. To characterize porosity, water absorption, and compressive modulus, GO scaffolds (SA-CS-Col-GO) were prepd. with and without Ca2+-mediated crosslinking (chem. crosslinking) and analyzed using Raman, Fourier transform IR (FTIR), X-ray diffraction (XRD), and SEM techniques. The incorporation of GO into the SA-CS-Col matrix increased both crosslinking d. as indicated by the redn. of cryst. peaks in the XRD patterns and polyelectrolyte ion complex as confirmed by FTIR. GO scaffolds showed increased mech. properties which were further increased for chem. crosslinked scaffolds. All scaffolds exhibited interconnected pores of 10-250 μm range. By increasing the crosslinking d. with Ca2+, a decrease in the porosity/swelling ratio was obsd. Moreover, the SA-CS-Col-GO scaffold with or without chem. crosslinking was more stable as compared to SA-CS or SA-CS-Col scaffolds when placed in aq. soln. To perform in vitro biochem. studies, mouse osteoblast cells were grown on various scaffolds and evaluated for cell proliferation by using MTT assay and mineralization and differentiation by alizarin red S staining. These measurements showed a significant increase for cells attached to the SA-CS-Col-GO scaffold compared to SA-CS or SA-CS-Col composites. However, chem. crosslinking of SA-CS-Col-GO showed no effect on the osteogenic ability of osteoblasts. These studies indicate the potential use of GO to prep. free SA-CS-Col scaffolds with preserved porous structure with elongated Col fibrils and that these composites, which are biocompatible and stable in a biol. medium, could be used for application in engineering bone tissues.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC1cXmt1Wmu7o%253D&md5=fba4604c1893657096bb29ef1fee8d06
45. 45
  Lee, K. Y.; Mooney, D. J. Alginate: Properties and biomedical applications. Prog. Polym. Sci. 2012, 37 (1), 106– 126, DOI: 10.1016/j.progpolymsci.2011.06.003
  45
  Alginate: Properties and biomedical applications
  Lee, Kuen Yong; Mooney, David J.
  Progress in Polymer Science (2012), 37 (1), 106-126CODEN: PRPSB8; ISSN:0079-6700. (Elsevier Ltd.)
  A review. Alginate is a biomaterial that has found numerous applications in biomedical science and engineering due to its favorable properties, including biocompatibility and ease of gelation. Alginate hydrogels were particularly attractive in wound healing, drug delivery, and tissue engineering applications to date, as these gels retain structural similarity to the extracellular matrixes in tissues and can be manipulated to play several crit. roles. This review will provide a comprehensive overview of general properties of alginate and its hydrogels, their biomedical applications, and suggest new perspectives for future studies with these polymers.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC3MXhsVaqt77L&md5=bfd5e6059760ff18f4ee316cc28fdb8b
46. 46
  Pellá, M. C.; Lima-Tenório, M. K.; Tenório-Neto, E. T.; Guilherme, M. R.; Muniz, E. C.; Rubira, A. F. Chitosan-based hydrogels: From preparation to biomedical applications. Carbohydr. Polym. 2018, 196, 233– 245, DOI: 10.1016/j.carbpol.2018.05.033
  46
  Chitosan-based hydrogels: From preparation to biomedical applications
  Pella, Michelly C. G.; Lima-Tenorio, Michele K.; Tenorio-Neto, Ernandes T.; Guilherme, Marcos R.; Muniz, Edvani C.; Rubira, Adley F.
  Carbohydrate Polymers (2018), 196 (), 233-245CODEN: CAPOD8; ISSN:0144-8617. (Elsevier Ltd.)
  A review. The advances in the field of biomaterials have led to several studies on alternative biocompatible devices and to their development focusing on their properties, benefits, limitations, and utilization of alternative resources. Due to their advantages like biocompatibility, biodegradability, and low cost, polysaccharides have been widely used in the development of hydrogels. Among the polysaccharides studied on hydrogels prepn., chitosan (pure or combined with natural/synthetic polymers) have been widely investigated for use in biomedical field. In view of potential applications of chitosan-based hydrogels, this review focuses on the most recent progress made with respect to prepn., properties, and their salient accomplishments for drug delivery and tissue engineering.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC1cXpvVKgtLk%253D&md5=4f0083dd48a76550cff3518e68a86104
47. 47
  Callahan, L. A. S. Combinatorial Method/High Throughput Strategies for Hydrogel Optimization in Tissue Engineering Applications. Gels 2016, 2 (2), 18 DOI: 10.3390/gels2020018
  There is no corresponding record for this reference.
48. 48
  Day, E. C.; Chittari, S. S.; Bogen, M. P.; Knight, A. S. Navigating the Expansive Landscapes of Soft Materials: A User Guide for High-Throughput Workflows. ACS Polym. Au 2023, 3 (6), 406– 427, DOI: 10.1021/acspolymersau.3c00025
  48
  Navigating the Expansive Landscapes of Soft Materials: A User Guide for High-throughput Workflows
  Day, Erin C.; Chittari, Supraja S.; Bogen, Matthew P.; Knight, Abigail S.
  ACS Polymers Au (2023), 3 (6), 406-427CODEN: APACCD; ISSN:2694-2453. (American Chemical Society)
  Synthetic polymers are highly customizable with tailored structures and functionality, yet this versatility generates challenges in the design of advanced materials due to the size and complexity of the design space. Thus, exploration and optimization of polymer properties using combinatorial libraries has become increasingly common, which requires careful selection of synthetic strategies, characterization techniques, and rapid processing workflows to obtain fundamental principles from these large data sets. Herein, we provide guidelines for strategic design of macromol. libraries and workflows to efficiently navigate these high-dimensional design spaces. We describe synthetic methods for multiple library sizes and structures as well as characterization methods to rapidly generate data sets, including tools that can be adapted from biol. workflows. We further highlight relevant insights from statistics and machine learning to aid in data featurization, representation, and anal. This perspective acts as a "user guide" for researchers interested in leveraging high-throughput screening toward the design of multifunctional polymers and predictive modeling of structure-property relationships in soft materials.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB3sXisFWgtrbF&md5=13654acd9ef4a5920a79b0f820a80a87
49. 49
  Yuan, X.; Liu, R.; Zhang, W.; Song, X.; Xu, L.; Zhao, Y.; Shang, L.; Zhang, J. Preparation of carboxylmethylchitosan and alginate blend membrane for diffusion-controlled release of diclofenac diethylamine. J. Mater. Sci. Technol. 2021, 63, 210– 215, DOI: 10.1016/j.jmst.2020.05.008
  49
  Preparation of carboxylmethylchitosan and alginate blend membrane for diffusion-controlled release of diclofenac diethylamine
  Yuan, Xiaoxue; Liu, Ran; Zhang, Wenchang; Song, Xiaoqiang; Xu, Lei; Zhao, Yan; Shang, Lei; Zhang, Jingsong
  Journal of Materials Science & Technology (Shenyang, China) (2021), 63 (), 210-215CODEN: JSCTEQ; ISSN:1005-0302. (Editorial Board of Journal of Materials Science & Technology)
  Controlled drug release technol. is becoming a com. sound methodol. of administering pharmaceutical therapies, and it is important to predict and control the release kinetics to fully take advantage of this technol. Carboxylmethylchitosan (CMCS) and alginate (SA) blend membranes were successfully prepd. by soln. casting technique and exposing to acetic acid atm., which induced the protonation of the amino groups of CMCS, and the formation of polyelectrolyte composite hydrogels. By using diclofenac diethylamine as the model drug, the diffusion controlled drug release behaviors of the membranes were studied based on permeation expt., which indicated that the swelling rates of the membranes could be adjusted by simply changing the wt. ratio of CMCS/SA and effectively changed the release rate of the membrane. The study shows that CMCS/SA blend membranes are promising biomaterials used for modulating the permeation behavior and developing advanced drug delivery systems.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BB38XhtFehsLzM&md5=6b7821ff6a3ccf5295435cfeb5580f03
50. 50
  Wu, P.; Fang, Y.; Chen, K.; Wu, M.; Zhang, W.; Wang, S.; Liu, D.; Gao, J.; Li, H.; Lv, J.; Zhao, Y. Study of double network hydrogels based on sodium methacrylate alginate and carboxymethyl chitosan. Eur. Polym. J. 2023, 194, 112137 DOI: 10.1016/j.eurpolymj.2023.112137
  There is no corresponding record for this reference.
51. 51
  Urayama, K.; Takigawa, T.; Masuda, T. Poisson’s ratio of poly(vinyl alcohol) gels. Macromolecules 1993, 26 (12), 3092– 3096, DOI: 10.1021/ma00064a016
  There is no corresponding record for this reference.
52. 52
  Gu, Y.; Zhao, J.; Johnson, J. A. Polymer Networks. In Macromolecular Engineering: From Precise Synthesis to Macroscopic Materials and Applications; Wiley-VCH, 2022; pp 1– 52.
  There is no corresponding record for this reference.
Supporting Information
Supporting Information
The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsami.4c16614.
- Additional methodological details of the fluid–structure interaction model; experimental results related to HTE studies, and a description of sample compositions, and Cantilever spectra; hydrogel photographs; raw sensor data, and summary of recent progress in literature (PDF)
- Representative data set for AE (XLSX)
- am4c16614_si_001.pdf (401.28 kb)
- am4c16614_si_002.xlsx (29.65 kb)
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.

Scalable Accelerated Materials Discovery of Sustainable Polysaccharide-Based Hydrogels by Autonomous Experimentation and Collaborative LearningClick to copy article linkArticle link copied!

ACS Applied Materials & Interfaces

License Summary*

Abstract

License Summary*

License Summary*

License Summary*

1. Introduction

2. Materials and Methods

2.1. Collaborative Learning via Consensus BO

2.1.1. Mathematical Notation

2.1.2. Independent Learning via Traditional BO

2.1.3. Collaborative Learning via Consensus BO─Model Description

2.2. Materials

2.3. High-Throughput Synthesis

2.4. High-Throughput Characterization

2.5. Scalable Self-Driving AMD via AE and Collaborative Learning

2.6. Statistical Analysis

3. Results and Discussion

3.1. Initializing Scalable Self-Driving AMD via AE and Collaborative Learning Using Labeled Data Acquired by HTE

Figure 1

Figure 2

Figure 3

3.2. Scalable AMD via AE and Collaborative Learning with Consensus BO

3.3. Comparison of Scalable Self-Driving AMD Performance via Collaborative vs Independent AE

Figure 4

4. Conclusions

Data Availability

Supporting Information

Terms & Conditions

Author Information

Acknowledgments

References

Cited By

ACS Applied Materials & Interfaces

License Summary*

Article Views

Altmetric

Citations

Recommended Articles

Abstract

Figure 1

Figure 2

Figure 3

Figure 4

References

Supporting Information

Supporting Information

Terms & Conditions

Scalable Accelerated Materials Discovery of Sustainable Polysaccharide-Based Hydrogels by Autonomous Experimentation and Collaborative Learning
Click to copy article linkArticle link copied!