Neighborhood Behavior:  A Useful Concept for Validation of “Molecular Diversity” Descriptors

Abstract

When searching for new leads, testing molecules that are too "similar" is wasteful, but when investigating a lead, testing molecules that are "similar" to the lead is efficient. Two questions then arise. Which are the molecular descriptors that should be "similar"? How much "similarity" is enough? These questions are answered by demonstrating that, if a molecular descriptor is to be a valid and useful measure of "similarity" in drug discovery, a plot of differences in its values vs differences in biological activities for a set of related molecules will exhibit a characteristic trapezoidal distribution enhancement, revealing a "neighborhood behavior" for the descriptor. Applying this finding to 20 datasets allows 11 molecular diversity descriptors to be ranked by their validity for compound library design. In order of increasing frequency of usefulness, these are random numbers = log P = MR = strain energy < connectivity indices < 2D fingerprints (whole molecule) = atom pairs = autocorrelation indices < steric CoMFA fields = 2D fingerprints (side chain only) = H-bonding CoMFA fields.

Dissimilarity-Based Algorithms for Selecting Structurally Diverse Sets of Compounds

“…A very simple, but important, task is that of comparing the many available methods to find those that are most suitable, both in terms of efficiency and effectiveness. Such studies are now appearing in the literature (Brown andMartin, 1996, 1997;Matter, 1997;Patterson et al, 1996) but only that by Snarey et al (1998) has provided a detailed analysis of some of the dissimilarity-based selection methods considered in this paper. More importantly, the full value of methods for analysis molecular diversity will only be obtained when they are linked to other, existing approaches to computer-aided molecular design, such as ligand docking, pharmacophore mapping and quantitative structure-activity relationships (Martin and Willett, 1998): the merits of such linked approaches are well illustrated by very recent work on the docking of combinatorial libraries (Jones et al, 1999;Kick et al, 1997), and we can expect many further such reports in the next few years.…”

Similarity‐based data mining in files of two‐dimensional chemical structures using fingerprint measures of molecular resemblance

“…As noted above when discussing cluster analysis, a cost-effective approach to biological testing involves a limited number of molecules that, taken together, describe the chemical space spanned by the complete set of molecules comprising a database (where this database could be an in-house corporate file, an external public or vendor database, or a set of molecules that could potentially be synthesised). The Similar Property Principle means that structurally similar molecules are likely to exhibit similar properties, and hence testing sets of similar molecules is unlikely to provide much more SAR information than would be obtained by testing just one or a few such molecules; instead, most information is likely to be obtained by testing sets of molecules that are as diverse, i.e., structurally dissimilar, as possible [80].…”

The EVA spectral descriptor

“…In terms of similarity/dissimilarity-based diversity analyses the Tripos neighbourhood behaviour criterion [32] provides a useful base-line from which to proceed. In essence a descriptor exhibits neighbourhood behaviour where small differences in a descriptor value tend to produce only a small difference in biological activity; i.e., high similarity in descriptor space implies similar biological activity.…”