Structure-based design of Mycobacterium tuberculosis PptT-ACP complex inhibitors using derivative design and machine learning.

Tuberculosis (TB), caused by Mycobacterium tuberculosis (MTB), remains a critical global health challenge, particularly with the rise of multidrug-resistant (MDR) strains. This study employed a comprehensive computational approach to identify and optimize inhibitors targeting the PptT-ACP complex, a key enzyme in MTB lipid biosynthesis. Virtual screening of FDA-approved compounds identified Mk3207 as a promising candidate. Stability analysis through molecular dynamics (MD) simulations validated its selection for derivative design. Three chemically tractable regions suitable for structural modification were identified, and the second region was selected for derivative design due to its favorable structural properties and binding interactions. One hundred derivatives were designed using ADMEopt and screened virtually, resulting in three top derivatives selected alongside Mk3207 for further evaluation. All compounds underwent 200 ns MD simulations in triplicate, with Compound_36 exhibiting the highest binding stability, as indicated by low root mean square deviation (RMSD) and root mean square fluctuation (RMSF) values, followed by Compound_98 and Compound_60. Free energy landscape (FEL) and principal component analysis (PCA) confirmed the thermodynamic stability of these derivatives. Predicted biological activity using machine learning (Random Forest Regression) indicated pIC50 values of 25.64, 22.43, 22.32, and 26.26 for Compound_36, Compound_98, Compound_60, and Mk3207, respectively. This study demonstrates the potential of derivative design and machine learning in designing potent MTB inhibitors, providing strong candidates for experimental validation to combat drug-resistant TB effectively.