Addressing the Metabolic Stability of Antituberculars through Machine Learning

Area of Science:

• Medicinal Chemistry
• Computational Chemistry
• Pharmacology

Background:

• Metabolic instability is a major challenge in drug discovery, particularly for tool compounds like antitubercular thienopyrimidines.
• The parent compound CD117 exhibits poor metabolic stability, with a mouse liver microsomal (MLM) half-life (t 1/2) under 1 minute.

Purpose of the Study:

• To prospectively apply and validate a novel Bayesian model for predicting MLM stability.
• To identify novel thienopyrimidine analogues with improved metabolic stability for antitubercular applications.

Main Methods:

• A virtual library of 411 analogues was generated by modifying the S-substituent of CD117.
• The MLM stability Bayesian model was used to prioritize 13 analogues for synthesis.
• Synthesized analogues were evaluated using MLM stability assays and biological profiling.

Main Results:

• All 13 synthesized analogues demonstrated superior metabolic stability compared to the parent compound.
• Six analogues achieved acceptable MLM t 1/2 values ( 2265 60 min).
• Simultaneous prediction of whole-cell efficacy and low mammalian cell cytotoxicity was not achieved.

Conclusions:

• The developed MLM stability Bayesian model is a valuable tool for optimizing chemical leads in drug discovery.
• The model successfully identified metabolically stable antitubercular thienopyrimidine analogues.
• Further development is needed to predict both efficacy and safety profiles concurrently.