SARgate: a structure–bioactivity analyser for navigating chemical space and mining SAR trends

This is the authors' abstract. We don't add key points for this paper.

The increasing availability of large chemical libraries and bioactivity datasets has created a growing need for cheminformatics tools capable of extracting interpretable structure-activity relationship (SAR) information across structurally diverse chemical series. However, many existing approaches rely on rigid scaffold definitions, descriptor-based clustering, or manually curated groupings, often limiting the identification of SAR trends spanning partially overlapping chemotypes. Here, we present SARgate, an open-source cheminformatics platform designed to organize chemical libraries into structurally coherent subsets and facilitate multi-level SAR exploration within a unified graphical environment. Starting from Bemis-Murcko scaffolds, SARgate applies automated aggregation procedures to derive generalized cores based on minimal shared substructures, enabling flexible R-group decomposition and improved recognition of chemically related series. Developed in Python using RDKit as the core cheminformatics engine, SARgate integrates dataset curation, scaffold organization, R-group analysis, matched molecular pair analysis (MMPA), stereochemical evaluation, similarity assessment, and structure-activity landscape visualization into a single interactive workflow accessible to users with different levels of computational expertise. The utility of SARgate is demonstrated through representative case studies involving AKT1 and IL4I1 inhibitor datasets derived from public repositories, externally curated collections, and manually assembled patent-derived libraries. These applications show that SARgate can recover known SAR determinants, identify activity-driving substituents and stereochemical constraints, reveal context-dependent effects, and support mechanistically interpretable medicinal chemistry insights directly from large-scale bioactivity data.