Privileged Structural Motif Detection and Analysis Using Generative Topographic Maps

Abstract

Identification of "privileged structural motifs" associated with specific target families is of particular importance for designing novel bioactive compounds. Here, we demonstrate that they can be extracted from a data distribution represented on a two-dimensional map obtained by Generative Topographic Mapping (GTM). In GTM, structurally related molecules are grouped together on the map. Zones of the map preferentially populated by target-specific compounds were delineated, which helped to capture common substructures on the basis of which these compounds were grouped together by GTM. Such privileged structural motifs were identified across three major target superfamilies including proteases, kinases, and G protein coupled receptors. Traditionally, the search for privileged structural motifs focused on scaffolds, whereas motifs were detected here without prior knowledge of compound classification in GTMs. This alternative way of navigating medicinal chemistry space further extends the classical, scaffold-centric approach. Importantly, detected motifs might also comprise fuzzy sets of similar scaffolds, pharmacophore-like patterns, or, by contrast, well-defined scaffolds with specific substituent patterns.

A Distance-Based Boolean Applicability Domain for Classification of High Throughput Screening Data

“…However, it should work in cases where people intend to exploit the chemical similarity principle . Or, as our peers would say, the data set must be neighborhood behavior-compliant , for the method to work.…”

AntiMalarial Mode of Action (AMMA) Database: Data Selection, Verification and Chemical Space Analysis

“…Molecules with the same RPs are considered to belong to a same cluster or “family”, because they share some common structural motif which is the underlying reason for their mapping into the same map zone “covered” by the RP. As already shown, common structural motifs may range from precisely defined scaffolds or even specifically substituted scaffolds, to fuzzier ensembles of related, interchangeable scaffolds, to even fuzzier “pharmacophore‐like” patterns.…”

Virtual Screening with Generative Topographic Maps: How Many Maps Are Required?

“…Projecting the multidimensional items (molecules for which high-dimensional descriptor elements each capture specific structural features) onto a two-dimensional (2D) latent grid is expected to mechanically reduce the predictive power, compared to any ideal machine learning method that operates in the original descriptor space. Nevertheless, previous studies ,,− have typically shown that GTM-driven classification or regression models are on par or only slightly less predictive than equivalent support vector machine or random forest approaches.…”

“…Like any global maps, their resolution is expectedly lower than the one that could be achieved by dedicated GTMs, focusing on specific series of compounds. The key question addressed in this work is whether such global maps, primarily conceived to serve as a rather coarse-grained “atlas” of the various structural motifs explored in to-date medicinal chemistry, ,, may nevertheless be successfully exploited as an accurate virtual screening and property prediction tool. This is envisaged by means of a consensus predictor using several universal maps, built on distinct initial descriptor spaces capturing distinct chemical information.…”

“…GTM was successfully used for structure−activity analysis of an antiviral compound set 8 and also of an antimalarial mode of action database. 9 Recently, it has also been successfully applied to visualize large public chemical databases such as PubChem, ChEMBL, 10 and FDB-. 11 Sidorov et al 12 applied GTM to create "universal" maps of chemical space that easily distinguished active and inactive compounds for more than 400 ChEMBL targets, yielding an averaged balanced accuracy (BA) higher than 0.6 for all targets, indicating high potential of this method for such applications.…”