Abhisek Jana, Aina Bellver-Sanchis, Koushik Barman, Christian Griñán-Ferré, Deb Ranjan Banerjee
Herein, we report a systematic ML-embedded virtual screening strategy for the repurposing of antifolates (hTS inhibitors) targeting Histone Lysine Methyltransferase (G9a/EHMT2), with experimental validation. Three pharmacophore models (3D QSAR, Common Feature, and Interaction), derived from known G9a inhibitors, were used in a database mining of the 1120 hTS inhibitors from the ChEMBL database. Candidates were filtered using a multi-tier screening strategy based on fit values, drug-likeness, ADMET, and TOPKAT criteria, followed by molecular docking to assess interactions with G9a's active site, including compounds such as UNC0642 and Raltitrexed as references. The top candidates were selected for strong docking scores, the ability to form at least three hydrogen bonds, and interaction with the lysine-binding tunnel. Among these, Pemetrexed exhibited a binding mode similar to Raltitrexed, forming critical hydrogen bonds and pi interactions with residues Leu1086, Ser1084, Tyr1154, Arg1157, and Phe1158, indicating its potential as a G9a inhibitor. Further analyses of RMSD, RMSF, ROG, and protein-ligand interactions confirmed the stability of Pemetrexed, and MM-GBSA binding energy calculations, along with residue-wise decomposition analysis, supported its strong affinity for G9a. An in vitro assay of Pemetrexed revealed G9a inhibitory activity with an IC50 of 304.90 nM, demonstrating its potency. Additionally, Pemetrexed exhibits lower activity against the G9a-like methyltransferase GLP than the references, confirming the compound's preferential activity for G9a over GLP. Furthermore, Pemetrexed improves the locomotor deficits associated with Aβ expression in C. elegans CL2006 worms in a dose-dependent manner, showing its potential as an epigenetic therapeutic for Alzheimer's Disease.