BY MARTIN SUMNER-SMITH

Collection and analysis of diverse information beyond DNA sequences is needed.

Timely selection of the most appropriate targets for therapeutic intervention is becoming the key determinant of future product success for pharmaceutical companies. Advances in genomics, proteomics, and related technologies have led to a wealth of potential new targets, but what processes can be implemented to determine the most attractive targets for therapeutic intervention? This question can be asked of the initiation step in many scientifically intense R&D project management environments.

Target selection requires the collection and analysis of a wealth of diverse information beyond DNA sequence and protein structure and role. It also requires mechanisms to facilitate process decisions. Fortunately, advances in knowledge and decision support technologies enable rigorous and effective business decision processes to be implemented. These processes can provide much-needed efficiency benefits to control the burgeoning costs of drug discovery and development. This article examines some of the information management aspects of target selection approaches being taken today by drug discovery organizations.

Structured information
Whereas initial target information is quite structured—it consists of a name and alternate names, DNA and protein sequences, protein structure, related genes, and so forth—a target selection decision must be based on a wide range of largely unstructured information, which is often derived from the published literature. Examples of questions that must be considered in selecting a target include:

• What is the normal physiological role of the protein?
• Does it have a pathological role?
• What diseases could intervention target?
• Are there other approaches to achieving the same therapeutic aim, and how do they compare?
• What is the potential for side effects?
• What is the patent and competitive landscape?
• What resources would be required?

The differences between structured and unstructured information present information management challenges, because it is usually not possible to rely on formal databases to manage all aspects of target selection. Summary annotation of candidate genes does not provide the necessary depth and context to support full assessment. The range of information required and the rapid implementation of new technologies demand a highly flexible environment. But they also require powerful indexing and retrieval methods to ensure that collected information is accessible.

Team input
Collecting and assessing the information needed to make a target-selection decision ideally require the input of a wide range of experts from different functional areas in a company. Target selection should not be the exclusive province of bioinformaticians, but it ought to include input from chemistry, manufacturing, business development, marketing, the legal department, and so forth. This input should be sought because others may have required professional knowledge, training, and expertise, and because they often have a better grasp of downstream discovery, development, and business requirements.

Because bioinformatics is a relatively new discipline in industry, most of its practitioners have limited industrial experience. Many joined companies with the expectation that they would develop novel algorithms or explore gene relationships without specific goals. In some companies at least, bioinformaticians are required to work in an applied fashion because the target-selection step has gained prominence as the beginning of the drug R&D process and as the first go versus no-go decision point. But ideally, corporate recognition of the importance of target selection should result in better team support and decision making, not just in a push to make bioinformaticians think in an applied manner.

Because input from so many individuals in different departments is sought, effective coordination is required. Each contributor must be informed when his or her input or review is required, and about the nature of that input.

In designing the repository, one must first decide what information to collect and analyze for each candidate target. Clearly, the information to be collected is based on an understanding of the information required to make an effective target-selection choice. This requirement can be mapped to a set of “folders”—either in a simple shared file system or, preferably, in a container-oriented knowledge management system. The advantage of a knowledge management system is that it provides version control, automated notification of changes, sophisticated permission control, and so forth, coupled with collaborative and reporting tools.

Hierarchical folder organization and naming conventions provide an intuitive representation that should be easy for any team member to understand. Examples of folders include links to DNA sequences, gene expression patterns in different tissues, physiological role, participation in defined pathways, potential therapeutic applications, and so on. Proper design and validation can simplify human navigation and assessment when information is presented in the same way for every target; but perhaps more importantly, it can also facilitate the subsequent direct comparison of targets.

The second, more challenging task is to decide who should collect each type of information, and when. If only a few targets are being reviewed, it may be sufficient for each team member to review each target in a repeating cycle, independently of the other team members. But if many targets are being reviewed, this is impractical, and computerized work-flow tools must be used. These allow the standardized design of a logical sequence of events for each target, both in series and parallel, with alternate paths taken on the basis of findings, and most importantly, with tailored real-time notification of each team member. Work-flow refinements include

• allowing team members to delegate,
• requiring input from one or more group members, and
• using automated processes for analysis (e.g., identifying new additions to databases).

Decision support
Once the necessary information has been collected, it must be provided in concise form to the appropriate decision makers. Ideally, this process minimizes the chances that a discovery project is launched that would generate a drug for which there is no market or whose introduction would ultimately be blocked by the patent position of competitors. Effective target selection can also lead to many other benefits, including improved chances of downstream success, less wasted effort, lower expenses, more effective resource allocation, and so forth.

Decision makers have access to information from many potential projects, but they need to be supported by query and reporting tools that allow them to track the status of each project and compare them in a variety of ways.

In the course of my work with different companies, I have seen several different target-selection approaches, each of which would benefit from the implementation of specific processes and decision support. However, most organizations have not considered formalized processes, perhaps because the necessary electronic support tools have not been available or used. Rather, there tend to be unstructured and unorganized efforts by teams that are driven more by scientific curiosity than specific goals.

Target selection
One target-selection approach is to first choose a particular class of genes and their products, such as a class of cell surface receptors, based on the historic suitability of that class as drug targets. Once this decision has been made, the next step is to review sequence databases and identify all possible candidates of the selected type. Because new gene family members may still remain to be identified, a subsequent process to review and add to the selected list should be planned. From this list, which may include a hundred or more candidates, the most promising targets are identified and advanced. However, this review must be iterative because the available information on most candidates is incomplete and constantly being added to and revised. We can expect that the review of potential targets will continue for many years. The paradigm shift that is under way in pharmaceutical discovery is a change from competitive advantage being derived from knowing of a target first, to using all the available information effectively to pick a promising target first and initiate a discovery project.

The target-selection process must include the input of many different experts, as discussed above. The best practice is to define the kinds of information that need to be collected on each target and the individuals responsible for each input. Managing both information collection and the process of notifying individuals of outstanding tasks requires an electronic environment, but it also requires the design and testing of the necessary processes. The information management advantage of this approach is that most of the information that must be passed on to the next stage (usually the development of screening assays) is already organized in electronic “folders”.

Another target-selection approach is to focus first on a specific disease, or at least diseases in specific therapeutic categories. Initial experiments focus on analyzing gene expression patterns, comparing tissues from various disease stages with normal tissue, and looking at the effects of current, effective drugs (if any) on gene expression. This process rapidly yields information on hundreds of genes. The subsequent process then focuses on whittling down the list to genes that seem central to the disease process, are comparatively unique to it, and whose products are likely to be amenable to therapeutic intervention. Generally, this approach requires the following two sets of “folders” keyed to project stages:

• a folder to collect information on the expression experiments, in which results on many genes are grouped according to tissue and experimental approach, and
• a subsequent folder set keyed to specific genes on which more in-depth information will be sought.

Summary
As the genomes of humans and pathogens become better described, target identification is becoming less a process of identifying novel targets and more one of identifying important characteristics about those targets before competitors do. The needs for rapid identification of these characteristics and implementation of subsequent discovery efforts make effective target selection the “front end” of the ongoing industrialization of drug discovery and development efforts. As the front end, target selection is also the ideal phase at which to implement knowledge management processes and support tools that can pass downstream for full project life-cycle management.

---

Martin Sumner-Smith is president and CEO of Base4 Inc. Send your comments or questions regarding this article to mdd@acs.org or the Editorial Office by fax at 202-776-8166 or by post at 1155 16th Street, NW; Washington, DC 20036.

 © 2001 American Chemical Society.  

---