Novel software tools for chemical diversity

Abstract

Although the concept of chemical diversity has been intuitively considered by chemists for many years, the advent of combinatorial chemistry and high-throughput screening have focused attention on the need for efficient software tools to address a variety of diversity-related tasks. Perhaps the most fundamental task related to chemical diversity is that of selecting a diverse subset of compounds from a much larger population of compounds. The obvious objective of that task is to identify a subset which best repre- sents the full range of chemical diversity present in the larger population, either to avoid the time and expense of synthesizing ‘redundant’ compounds or to avoid the time and expense of screening ‘redundant’ compounds. However, in addition to simple subset selection, recent practical experience has revealed other equally (possibly more) import- ant diversity-related tasks which must also be addressed in the pharmaceutical and agrochemical industry. High-throughput screening (HTS) can be an effective approach to lead discovery but is obviously limited by the structural diversity of compounds being screened. What if that population does not include representatives of one or more chemical classes or pharmacophores? Identifying diversity-voids or missing diversity is an important task and, obviously, filling in diversity voids with compounds from other sources is equally important. It is also important to be able to recognize and choose among the many com- pounds which might fill a particular diversity-void. These tasks become increasingly important as the number of combinatorially synthesizable compounds increases with advances in combinatorial chemical methods, and as the number of commercially avail- able compounds increases. Similarly, comparing diversities of alternative compound libraries is another important diversity-related task. In addition to simple diverse subset selection, it is often desirable to select a subset chosen not only to provide structural diversity, but also to satisfy one or more non- structural criteria or ‘biases’. For example, compound availability and/or physical prop- erties may be important when selecting a subset for HTS purposes. Reagent cost and/or reagent usage frequency may be important when deciding which compounds actually to synthesize out of a very large range of compounds synthetically accessible through combinatorial chemical methods. Clearly, non-structurally biased subset selection will yield subsets with somewhat less structural diversity than a subset chosen simply to maximize structural diversity, but practical considerations often make biased subset selection a very important diversity-related task. The following section will discuss how chemical structures can be described for chemical diversity purposes. We shall refer to such descriptors as ‘metrics’ of a ‘chemistry- space’. An extremely important but often overlooked diversity-related task is that of choosing the chemistry-space metrics which best represent the structural diversity of a given population of compounds. For example, combinatorially generated populations, or populations chosen to be similar to a particular active (‘lead’) compound, are inherently less diverse than other more randomly assembled populations. Thus, it is quite reason- able to expect that metrics specifically tailored to focus on the limited diversity of such ‘focused populations’ will provide some advantages over metrics which were developed to best represent the broad range of diversity found in ‘non-focused populations’. Last but not least, the notion of considering alternative chemistry-space metrics reminds us of the need for a rational approach for validating chemistry-space metrics — an important and often misunderstood diversity-related task.