Reaction-Based De Novo Design of Synthetically Accessible Competitive Inhibitors Targeting Glutathione S-Transferase Theta 1 (GSTT1)

GSTT1 (glutathione S-transferase theta 1) is essential for metastatic tumour cell dissemination, making it a promising anti-cancer target. This study employed reaction-based de novo design to generate novel competitive GSTT1 inhibitors optimised for both target binding and synthetic accessibility. Using the Synopsis software and two commercial libraries (Chemspace and Enamine), eight distinct inhibitors were computationally generated, each with a defined synthetic route comprising one to four steps. Molecular docking simulations using AutoDock Vina predicted strong binding free energy (BFE) for all inhibitors (-10.1 kcal/mol to -11.2 kcal/mol). Together with the structural analysis showing the presence of distinct polar and nonpolar regions in each inhibitor, this suggests effective dual-site engagement with both the hydrophilic glutathione-binding (G-site) and hydrophobic substrate-binding (H-site) pockets of the GSTT1 active site. In silico pharmacokinetic profiling using SwissADME further suggested that the de novo GSTT1 inhibitors had favourable oral pharmacokinetics and drug-like properties, although some were predicted to be P-glycoprotein (P-gp) substrates, inhibitors of major cytochrome P450 (CYP450) isoforms, or capable of permeating the blood-brain barrier (BBB). Overall, three top candidates (Syn-C1, Syn-C2, Syn-E3) emerged with the best balance of affinity, synthetic feasibility, and pharmacokinetic profiles. A key limitation of this study is its reliance on a static protein structure, which may not fully capture the conformational dynamics of GSTT1. Future work should address this by incorporating molecular dynamics simulations to generate conformational ensembles of GSTT1, thereby improving the physiological relevance of binding predictions. De novo inhibitors could also be further optimised using open-source tools such as AutoGrow4, while machine learning-based in silico ADME prediction methods could provide more accurate pharmacokinetic profiling. While this study highlights the immense potential of computational workflows to accelerate drug design, experimental approaches remain essential to validate in silico findings.