Rational Screening Set Design and Compound Selection:  Cascaded Clustering

Abstract

The use of cascaded clustering is reported. This technique was developed to permit the application of Jarvis-Patrick clustering based on structural fingerprints to large chemical databases, while keeping the maximum cluster size and the number of singletons produced at reasonable levels. The basis for the algorithm, its implementation, and validation are described. In the first part of the paper, the approach is used to create a representative subset of compounds for biological testing from the corporate compound repository. A variation of the method is then used for the comparison of relatively large databases. Finally, compound selection using cascaded clustering is shown to be complementary to the Diverse Property-Derived approach, which is based on partitioning by six molecular descriptors.

Balancing Representativeness Against Diversity using Optimizable K-Dissimilarity and Hierarchical Clustering

“…This option (recursive OptiSim clustering) may be the alternative of choice when working with heterogeneously distributed populations for which the extent of structural variation is misleadingly different in different regions of the structural space; an analogous approach has recently been proposed for improving the distributions obtained from Jarvis-Patrick clustering. 13 Interaction between Subset Size M and Subsample Size K. The results described above were obtained for a relatively small population (N ) 1000) and for a fixed subset size (M ) 30). Figure 5 shows the results obtained when OptiSim was used to select from 5 to 50 compounds from the full mixed library of 6600 compounds.…”

Status of HTS Data Mining Approaches

“…One is to add the obtained singletons iteratively to the clusters in a fuzzy way [52]. Another approach is to iteratively remove singletons and re-cluster them using milder Jarvis-Patrick conditions [53]. Irrespective of these developments new methods are introduced continuously [54 ± 56].…”

Using High-Throughput Screening Data To Discriminate Compounds with Single-Target Effects from Those with Side Effects

“…[28][29][30][31][32][33][34][35] In a notable exception, decision trees were used to model activity on a small number of different targets simultaneously. 36 Alternatively, clustering algorithms have been widely used to identify distinct biologically active compound classes [37][38][39][40] and to segregate compounds into distinct structural classes that differ in their assay phenotypes. 41 This paper explores the potential of computational tools, namely clustering and an entropy-based coincidence score, to identify compound classes active in multiple assays and generate hypotheses about a compound's biological target(s) or mechanism(s) of action.…”