Machine learning and docking models for Mycobacterium tuberculosis topoisomerase I

Area of Science:

• Medicinal Chemistry
• Drug Discovery
• Computational Biology

Background:

• There is a critical need for new drugs against Mycobacterium tuberculosis, as existing treatments face resistance and target limited pathways.
• Mycobacterium tuberculosis Topoisomerase I (Mttopo I) is an essential enzyme and a promising, yet underexplored, target for novel anti-tubercular therapies.
• A scarcity of known Mttopo I inhibitors necessitates innovative approaches for identifying potential drug candidates.

Purpose of the Study:

• To identify novel small molecule inhibitors of Mttopo I.
• To develop and validate machine learning models for predicting Mttopo I inhibitors.
• To guide the selection of compounds for in vitro screening and experimental validation.

Main Methods:

• Computational methods including homology modeling and molecular docking were employed.
• A library of 639 compounds was screened in vitro for Mttopo I inhibition.
• Machine learning models, specifically a Bayesian model, were developed and validated using screening data.

Main Results:

• Machine learning models demonstrated good predictive performance (5-fold cross-validation ROC of 0.74, sensitivity/specificity/concordance > 0.76).
• The validated models were used to select commercially available compounds for further in vitro testing.
• Experimental inhibition of Mttopo I by small molecules was confirmed, validating the enzyme as a druggable target.

Conclusions:

• A multi-faceted approach combining machine learning, docking, and experimental screening is effective for identifying Mttopo I inhibitors.
• This study successfully identified small molecule inhibitors of Mttopo I, providing a foundation for lead molecule development.
• Targeting Mttopo I represents a viable strategy for developing new anti-tubercular agents.