Machine Learning-Driven Discovery of Indole/Oxoindole–Piperazine Scaffolds as Dual MAO-B/Sig-1R Ligands for Neurodegenerative Disorders

-aryl piperazines), as a working hypothesis for multitarget MAO-B/Sig-1R modulation. A ligand-based machine learning approach was applied to build robust quantitative structure-activity relationship (QSAR) classifiers capable of identifying potential modulators of both targets. Using curated ChEMBL data sets, support vector machine models were trained with molecular descriptors selected via L1 regularization and forward feature selection. The final models achieved strong predictive performance (κ > 0.78; accuracy >88%) and revealed key structural determinants of activity, including aromatic bridging/planarity, polarizability-related terms, and complementary donor-acceptor patterns. Virtual screening of 147 natural compounds using the developed QSAR models, followed by applicability domain filtering, identified 11 compounds with predicted dual-target potential. Subsequent docking-based analysis prioritized four compounds with plausible predicted binding modes at both targets, while molecular dynamics simulations (150 ns) supported the short-time scale stability of 2 selected protein-ligand complexes. Overall, this work establishes general SAR principles for dual MAO-B/Sig-1R modulation and highlights how descriptor-driven modeling can inform molecular design and the prioritization of compounds for synthesis in multitarget medicinal chemistry programs.