Molecular Diversity and Representativity in Chemical Databases

Abstract

It is now common practice in the pharmaceutical industry to use molecular diversity selection methods. With the advent of high throughput screening and combinatorial chemistry, compounds must be rationally selected from databases of hundreds of thousands of compounds to be tested for several biological activities. We explore the differences between diversity and representativity. Validation runs were made for different diversity selection methods (such as the MaxMin function), several representativity techniques (selection of compounds closest to centroids of clusters, Kohonen neural networks, nonlinear scaling of descriptor values), and various types of descriptors (topological and 3D fingerprints) including some validated whole-molecule numerical descriptors that were chosen for their correlation with biological activities. We find that only clustering based on fingerprints or on whole-molecule descriptors gives results consistently superior to random selection in extracting a diverse set of activities from a file with potential drug molecules. The results further indicate that clustering selection from fingerprints is biased toward small molecules, a behavior that might partly explain its success over other types of methods. Using numerical descriptors instead of fingerprints removes this bias without penalising performance too much.

Effect of Molecular Descriptor Feature Selection in Support Vector Machine Classification of Pharmacokinetic and Toxicological Properties of Chemical Agents

“…For instance, descriptors for the QSAR of relatively small sets of related agents are not applicable for the analysis of chemical diversity. 22 The use of descriptors unrelated to a particular type of properties or biological activity likely generates noise in a statistical learning system, which may affect the prediction accuracy of that system. 22 In some cases, it is difficult to manually select descriptors useful for a particular property.…”

“…22 The use of descriptors unrelated to a particular type of properties or biological activity likely generates noise in a statistical learning system, which may affect the prediction accuracy of that system. 22 In some cases, it is difficult to manually select descriptors useful for a particular property. Thus methods capable of automatic selection of molecular descriptors are desirable.…”

Assessing the Chemical and Biological Diversity of an Ion Channels Knowledge Database

“…Subsequently, a principal component analysis (PCA) study 23,24 was performed using 12 GCUT bidimensional molecular descriptors implemented in MOE 15 to represent the chemical space covered by the ligand of a given target family. Among these descriptors, three were found to be correlated, and were subsequently excluded.…”

Grouping of Coefficients for the Calculation of Inter-Molecular Similarity and Dissimilarity using 2D Fragment Bit-Strings

“…In this paper, we focus on similarity coefficients, in particular their use for measuring similarities between pairs of 2D fragment bit-strings (or fingerprints). These representations provide a very simple encoding of molecular structure but have been found to yield a surprisingly high level of performance in a range of similarity and diversity studies [2,[6][7][8][9][10].…”