Optimization and visualization of molecular diversity of combinatorial libraries

Abstract

One of the major goals of rational design of combinatorial libraries is to design libraries with maximum diversity to enhance the potential of finding active compounds in the initial rounds of high-throughput screening programs. We present strategies to visualize and optimize the structural diversity of sets of molecules, which can be either potential substituents to be attached at specific positions of the library scaffold, or entire molecules corresponding to enumerated libraries. The selection of highly diverse subsets of molecules from the library is based on the stochastic optimization of 'Diversity' functions using a single-point-mutation Monte Carlo technique. The Diversity functions are defined in terms of the distances among molecules in multidimensional property space resulting from the calculation of 2D and 3D molecular descriptors. Several Diversity functions, including an implementation of D-Optimal design, are applied to select diverse subsets and the results are compared. The diversity of the selected subsets of molecules is visualized by embedding the intermolecular distances, defined by the molecules in multidimensional property space, into a three-dimensional space.

Physical interaction of apoptosis-inducing factor with DNA and RNA

“…The different oligonucleotides were docked to AIF with QXP (McMartin and Bohacek, 1997), as outlined in Table 1. Analysis and selection of AIF/RNA complexes was based on scoring, clustering and visual inspection (Table 1), using published methods (Hassan et al, 1996;Giordanetto et al, 2004).…”

Dissimilarity-Based Algorithms for Selecting Structurally Diverse Sets of Compounds

“…In this work, molecules are represented by principal components derived from calculated physical properties (topological and information content indices, and electronic, hydrophobic and steric descriptors) (Hassan et al, 1996) or by low-dimensionality autocorrelation vectors describing the distribution of the electrostatic potential over the van der Waals' surface of a molecule (Agrafiotis, 1997), and the scoring function for the SA uses one of several different inter-molecular distance functions in the resulting descriptor space. Another example of the use of SA asa searching tool is provided by the HARPick program (Good and Lewis, 1997).…”

ChemTreeMap: an interactive map of biochemical similarity in molecular datasets

“…With MiniBatch-Kmeans and RDKit fingerprint, the clustering step for ChEMBL-the largest dataset-took <5 h to run on a machine with 16 Gb memory (i3-2100 CPU @ 3.10 GHz). In comparison, the maximum dissimilarity method (Hassan et al, 1996) implemented in PipelinePilot (Hassan et al, 2006) takes more than a week to cluster ChEMBL, and an algorithm with average runtime complexity O(N 3 log N) such as DBScan still takes more than 3 days.…”