Sorting Out Combinatorial Chaos Software helps keep track of the thousands of compounds created by parallel synthesis. BY NICHOLAS SLEEP For a long time, chemists in search of new drugs created compounds one at a time. Using medicinal chemistry principles and intuition, researchers would create a molecule and then study it extensively. After the scientists learned which properties they liked or did not like about the molecule, they would synthesize a variation and study it. This method resulted in an increased understanding of many biological systems, but it has been perceived to be a slow way to find new drugs. When the idea of parallel synthesis first entered the realm of drug discovery, it was seen as a technique that could yield vast chemical diversity. By covering the range of chemical diversity, scientists would find “hits”—biologically relevant molecules that might become drugs—faster. Unfortunately, because there were more drugs to screen, more resources had to be spent on assays. Sheer unrestricted diversity in the chemical libraries did not yield the percentage of hits that researchers dreamed of. Questions emerged about the cost-effectiveness of combinatorial chemistry in drug discovery. Researchers quickly realized that sampling the entire range of chemical diversity was not necessarily appropriate for picking compounds with biological targets. They started to create more focused compound libraries based on principles learned from biology. The resulting compounds sample the druglike space in the diversity continuum. Compound libraries used to screen against a new biological target may sample a large portion of this space; smaller libraries might be used for better-known targets. Lead optimization can also harness the power of combinatorial chemistry by creating variations on initial lead compounds to create molecules that bind better, have more target specificity, or have better bioavailability. These focused libraries represent a valuable resource for pharmaceutical companies. Of a large set of compounds created for a particular target, only a very few will be pursued for that target. But the remainder of the compounds are still druglike and might be useful for another, yet-to-be-identified target. So instead of recreating the set of compounds each time, most companies retain their libraries to get the best return on their investment. Keeping up with the compounds As combinatorial chemistry increases the size of a company’s main compound library, the complexity of managing these libraries increases as well. But for the library to remain useful to the R&D process, it needs to be accessible. Not only is it essential to manage the repository to physically locate and handle the compounds, but it is also vital to maintain detailed records of sample use and data from previous assays, while conserving stocks by restricting the use of scarce compounds. The process, results, and materials must be managed in an integrated fashion, from combinatorial chemistry, through high-throughput screening, to development and lead optimization. Most companies begin with a manual repository system. After a compound is created and analyzed, it is labeled and stored manually in a central location. Information technology systems can help scientists access compound information and analyze chemicals and biological data. Compound retrieval is simply a matter of going to the storage area, reaching for the compound, and returning with it to the bench. No records are kept of compound usage or stock levels. But as the compound collection grows, sample retrieval and preparation become increasingly unwieldy. With a larger library, more compounds will be screened against targets, and the number of targets is likely to increase as researchers take advantage of the capabilities of high-throughput screening. Automated liquid handling and robotic storage and retrieval systems are usually introduced at this stage together with automated chemical synthesis equipment for library generation and lead optimization. At this time, careful consideration must be paid to their associated information-handling systems. Compounds can be lost in a large system—created and stored but never retrieved. The fear, of course, is that the next big blockbuster drug (or at least the lead compound that will be shaped into a blockbuster) is “gathering dust” in the library. Companies often opt for an ad hoc solution, writing additional software to add inventory-management capability to one of their existing applications. Although this approach can work up to a point, it often creates a highly complex software environment. Complexity increases the potential for software bugs, increases development time and costs, and offers users limited flexibility as needs and process demands change. For example, a company may decide to solubilize a compound in multiple solvents where previously they had used a single solvent. This decision could have significant effects on both the business logic and the core schema of a company’s inventory management system (IMS). Another alternative is to buy inventory-control software from a vendor specializing in this sort of system. This approach can certainly reduce the time and costs of in-house software development. Third-party inventory software, however, often duplicates the process-control functionality provided with automated liquid handling and robotic storage and retrieval systems. Duplication can lead to confusion for operators and suboptimal process flows. The go-between A third approach, one that gets around many of these problems, relies on a three-tier architecture that sandwiches new software between existing systems (Figure 1). In this arrangement, an IMS sits between the company’s existing inventory system(s) and the automation vendor’s process-control software. The IMS provides a detailed view of the available inventory and serves as a central clearinghouse for orders, recording requests and arbitrating between multiple requests for the same compound. The software dispatches dispensary requests from the company’s existing compound request systems (CRSs) to the process management/robotics control (PM/RC) system, tracks the results of PM/RC activity, and updates the inventory accordingly. Figure 1. The IMS enables the customer’s existing compound request systems to interface with the automation vendor’s process management and robotics control software by acting as a central clearinghouse for orders. The repository manager also uses this system to control and oversee the repository. A good IMS should take advantage of this architecture by providing additional inventory-control functionality. As these features are required only by repository staff, they need not reside in the CRS that the researchers see, so the expense of modifying the CRS is reduced. Examples of new functions include approving or modifying orders placed by scientists at the CRS and stock control—that is, disposing of unneeded compounds, replenishing liquid stock by replacing a tube, and placing plate production orders. These requirements are relatively standard throughout the pharmaceutical industry. Because every order goes through the IMS, it can implement a lock-and-reservation system to limit access to particular substances or vessels to a specified team or group of individuals. This type of security system increases scientists’ faith in the repository because it ensures that no one else can use that compound without their knowledge and permission. As all orders pass through the IMS, it can keep detailed audit trails of compound usage. A single audit trail follows each compound from its initial registration to the point when it enters the high-throughput assay tracking system. This historical information can be used to correlate results between screens and help the repository manager spot trends in compound usage. As closed-loop screening systems develop with the IMS at their heart, this single audit trail will become increasingly valuable. Closed-loop screening Automation and information handling have become increasingly important not only for primary screening of large compound libraries, but also for secondary and tertiary screening. Secondary screening confirms that active compounds respond the same way during retesting as they did in the initial screen. In other words, the compounds are real hits and have not arisen as a result of systematic errors from human or automated systems. Compounds whose activity is confirmed on secondary screening are submitted to a tertiary screen in which the IC50 (the concentration at which a compound is half maximally active) is determined. This screen allows the relative potencies of different hits to be compared. Fast and accurate sample processing is essential for closed-loop screening applications (Figure 2), in which cherry-picking and follow-up screening of active samples begin even before primary screening is completed. Figure 2. Closed-loop screening systems have evolved from inventory management systems to eliminate systematic errors and ensure the validity of “hits”. To date, there have often been significant delays between initial identification of the primary screening hit and subsequent passes through secondary and tertiary screens. These interprocess delays affect the reproducibility of subsequent screens as additional variables are introduced. Variables include the screen being performed by different operators on different assay platforms, often with batches of reagents different from those used in the initial primary screen. Potentially, even the screening method could change. If these variables are introduced into the biological tests, they can lead to discrepancies in interpreting the data. By coupling primary and subsequent screens, many variables are eliminated. In addition, the desire of the sponsoring organization to progress from screen hits to confirmed hits with associated potency data is hampered by the long delays that are currently encountered. Until these data are obtained, it is not practical to progress to the later stages of lead optimization, which include selectivity, toxicity testing, and the generation of hit analogues by medicinal chemistry. In an increasingly competitive field, these delays can potentially affect a company’s competitive position. In short, closed-loop screening offers the possibility of generating better-quality data in less time, which results in an increased competitive advantage. Consolidating inventory management The Automation Partnership (TAP) developed its IMS to integrate with Haystack, the company’s automated sample storage and retrieval system. Individual users can link their existing CRSs with TAP’s IMS, which can control various process management and robotics systems. TAP designed its initial IMS, known as the Compound Ordering System, specifically for Haystack; variations of the IMS now in development will allow for integration with other storage and retrieval systems. One Haystack and IMS system in development will include features designed to meet the needs of combinatorial chemistry laboratories. TAP’s IMS is a Web-based system that scientists and managers install on their desktop computers without the need for additional software. One of the virtues of Web technology is that it provides multisite access to the system over the company’s intranet. There is no technical reason why companies could not allow their collaborators to order directly over the Web to streamline the research process. The IMS generates a single inventory database with which other linked software systems can interact through a series of ordering screens. In addition to creating audit trails and tracking the amount of inventory for each compound—whether it is stored in bottles, tubes, or plates—the IMS uses algorithms that allow it to create reservations and locks on specific substances or types of vessels. For example, a repository group may have access only to substances stored in tubes, and the IMS would deny requests for compounds stored in bottles or plates. Engineering a solution As combinatorial chemical synthesis and high-throughput screening systems advance in speed and complexity, they will necessitate not only more widespread, integrated automation of critical processes, but, on a broader level, industrialization of the overall drug discovery process. Production-scale compound synthesis and screening operations will involve storage, tracking, and retrieval of perhaps hundreds of millions of compounds. To take the fullest advantage of new technologies, this industrial-scale process will demand database management solutions that are flexible, accessible, expandable, and secure. As large pharmaceutical companies seek ways to bring new drugs to the market more quickly, they may model their efforts on the concept of the drug discovery factory (see “Drug discovery adopts factory model,” Modern Drug Discovery May 2000, p 67). In this model, the factory represents an industrial-scale production facility that embodies the parallel compound synthesis and high-throughput screening activities of the pharmaceutical company. The model maintains both a physical and philosophical separation between these highly automated, large-scale operations and the company’s intellectually driven research activities by allowing the company’s scientists to concentrate on generating ideas while passing the task of turning these demands into action to the factory’s specialists in production and batch processing. The concept of the drug discovery factory will evolve as the technologies evolve. The convergence of process automation and software engineering at many different levels will drive much of the refinement in the model. One such convergence has created the need for the IMS. The IMS, in turn, will influence the course of the drug discovery factory. Nicholas Sleep is a product manager with The Automation Partnership. Comments and questions for the author may be addressed to the Editorial Office by e-mail at mdd@acs.org, by fax at 202-776-8166 or by post at 1155 16th Street, NW; Washington, DC 20036. Top || Modern Drug Discovery Home Page