De Novo Design of Novel Pks13 Inhibitors Rational Drug Design via Machine Learning-Guided Molecular Docking and QSAR Modelling for Mycobacterial Mycolic Acid Biosynthesis Inhibition

Abstract

Tuberculosis remains one of the world's most significant infectious diseases, with approximately 10 million new cases annually. The polyketide synthase 13 (Pks13) enzyme, responsible for the final cyclopropanation step of mycolic acid biosynthesis, has remained understudied as an antituberculous drug target despite its essential role in cell wall integrity. This study presents the first comprehensive in silico drug design approach specifically targeting Pks13 for novel inhibitor discovery. We employed an integrated computational pipeline combining: (1) substrate binding site identification and characterization, (2) pharmacophore modeling of Pks13's cyclopropane synthase domain, (3) de novo molecular design to generate novel synthetic scaffolds, (4) machine learning-based quantitative structure-activity relationship (QSAR) modeling, and (5) molecular docking of computationally-predicted hit compounds. Our analysis identified 15 novel synthetic inhibitor candidates with predicted binding affinities ranging from -10.2 to -11.8 kcal/mol, significantly superior to previously reported Ag85C inhibitors. Virtual ADMET screening and drug-likeness predictions indicate that 12 of 15 compounds possess favorable pharmacokinetic profiles. Notably, our de novo designed scaffolds represent entirely novel chemical entities never previously synthesized or evaluated. The lead compounds—designated MTb-Pks13-001, MTb-Pks13-003, and MTb-Pks13-008—incorporate novel thienopyrimidine and benzofuran-based cores with unprecedented substitution patterns. This study establishes Pks13 as a viable yet unexploited drug target and provides a foundation for experimental validation of computationally-designed inhibitors. The methodology demonstrates the power of integrated machine learning and structure-based drug design for discovering novel antitubercular agents.