Validating new tuberculosis computational models with public whole cell screening aerobic activity datasets.

Area of Science:

• Drug discovery and development
• Computational chemistry
• Infectious diseases

Background:

• High-throughput screening and computational methods are crucial for identifying new anti-tuberculosis agents.
• Public datasets from the Collaborative Drug Discovery Tuberculosis (CDD TB) database are valuable resources for drug discovery.
• Cheminformatics approaches can validate dataset utility and suggest compounds for testing against Mycobacterium tuberculosis (Mtb).

Purpose of the Study:

• To evaluate the utility of public tuberculosis datasets using cheminformatics.
• To identify potential drug leads for tuberculosis using computational methods.
• To validate Bayesian classification models and SMARTS alerts for drug discovery.

Main Methods:

• Bayesian classification models were employed to predict activity against Mtb in a set of Novartis compounds.
• FDA-approved drugs were screened using computational models.
• Novartis compounds were filtered using computational SMARTS alerts to identify undesirable substructures.

Main Results:

• Bayesian models showed >4.0-fold enrichment for aerobic hits and 10-fold enrichment for Mtb-active compounds in FDA drugs.
• 85.9% of Novartis compounds failed SMARTS alerts, a higher rate than known TB drugs.
• SMARTS filter failures correlated with Lipinski rule violations, indicating potential drug-likeness issues.

Conclusions:

• Computational approaches, including Bayesian models and SMARTS filtering, can aid in identifying desirable leads for tuberculosis drug discovery.
• These methods offer an efficient way to prioritize compounds for further testing.
• The study highlights the potential of cheminformatics in accelerating the development of new anti-TB drugs.