Neighborhood-Based Prediction of Novel Active Compounds from SAR Matrices

Abstract

The SAR matrix data structure organizes compound data sets according to structurally analogous matching molecular series in a format reminiscent of conventional R-group tables. An intrinsic feature of SAR matrices is that they contain many virtual compounds that represent unexplored combinations of core structures and substituents extracted from compound data sets on the basis of the matched molecular pair formalism. These virtual compounds are candidates for further exploration but are difficult, if not impossible to prioritize on the basis of visual inspection of multiple SAR matrices. Therefore, we introduce herein a compound neighborhood concept as an extension of the SAR matrix data structure that makes it possible to identify preferred virtual compounds for further analysis. On the basis of well-defined compound neighborhoods, the potency of virtual compounds can be predicted by considering individual contributions of core structures and substituents from neighbors. In extensive benchmark studies, virtual compounds have been prioritized in different data sets on the basis of multiple neighborhoods yielding accurate potency predictions.

The ‘SAR Matrix’ method and its extensions for applications in medicinal chemistry and chemogenomics

“…Thus, from NBHs, "mini-QSAR" models are derived for activity prediction. For each candidate VC, qualifying NBHs are collected across all SARMs, individual potency predictions are carried out, and their consistency is evaluated, for example, by calculating standard deviations for predictions 5 . In benchmark calculations on six different sets of G protein-coupled receptor ligands, potency values of subsets of test compounds falling into continuous local SAR regions were accurately predicted using the NBH-based approach, and prediction accuracy generally increased with the number of qualifying NBHs 5 .…”

“…The Structure-Activity Relationship Matrix (SARM) approach has originally been designed to extract and organize SAR-informative compound series from large data sets 4 and has been further extended to help bridge the gap between data-driven SAR analysis, compound design, and activity predictions 5 and study compound series in multi-target activity spaces 6 . Here, we present the SARM approach and its extensions in context and introduce new features and applications.…”

A Novel Fragment Recommendation Workflow using Direct and Indirect Transfer of SAR According to Integrated Similarities of Scaffold Motifs and SAR Trends: Application to Identifying Factor Xa Inhibitors

“…The Matsy algorithm searches a series of fragments present in SAR databases according to the rank order of activity profiles and proposes alternative fragments [13]. Bajorath reported that matrices of SAR information were used to estimate profiles of newly designed compounds through summing the activity profiles of related neighborhood compounds in the matrices [14]. However, the Matsy procedure only detects a fragment series that is directly related to the series of query fragments, and Bajorath's method focuses on their scaffolds of MMP relationships.…”

Hit Expansion from Screening Data Based upon Conditional Probabilities of Activity Derived from SAR Matrices

“…[10] Because SARMs contain both data set compounds with known activity and virtual compounds, matrix neighborhoods of virtual compounds that exclusively consist of known active compounds can be systematically identified and neighborhood information can be used for activity prediction. [11] In SARMs, compounds are represented as coresubstituent combinations. Therefore, the activity of virtual Abstract: A new methodology for activity prediction of compounds from SAR matrices is introduced that is based upon conditional probabilities of activity.…”

“…Accordingly, a 'mini Free-Wilson model' was built for each qualifying SARM neighborhood of a given virtual compound to predict its activity. [11] Virtual compounds often have multiple qualifying neighborhoods in SARMs, which makes it possible to assess the consistency of activity predictions.…”