MtbAIM: AI/ML-guided endpoint-specific prediction of small-molecule activity across drug-susceptible and drug-resistant Mycobacterium tuberculosis

AI/ML workflow predicts small-molecule activity against drug-susceptible and drug-resistant Mycobacterium tuberculosis with high accuracy, achieving AUROC scores above 0.83.

• Why it matters: Addressing drug resistance in tuberculosis is crucial because compounds effective against susceptible strains often fail against resistant ones, hindering drug development and treatment strategies.
• What they did: The study developed an AI/ML approach using diverse molecular representations and scaffold-held-out validation, analyzing over 29,000 compounds to model MIC and related endpoints separately for DS and DR strains.
• The result: This method enables precise prioritization of compounds for testing against both strain types, and the MtbAIM platform makes these predictions accessible via Google Colab, supporting tuberculosis drug discovery efforts.