Machine Learning-Based High-Throughput Screening, Molecular Modeling and Quantum Chemical Analysis to Investigate...

Area of Science:

• Computational chemistry
• Drug discovery
• Tuberculosis research

Background:

• Drug resistance in Mycobacterium tuberculosis (Mtb) poses a significant treatment challenge.
• Existing therapies for tuberculosis are often limited by severe side effects and emerging resistance.
• Novel therapeutic strategies targeting essential Mtb enzymes are urgently needed.

Purpose of the Study:

• To identify novel drug candidates targeting Mtb methionyl-tRNA synthetase (MtbMetRS) using in silico approaches.
• To screen a large library of molecules for potential MtbMetRS inhibitors.
• To evaluate the binding affinity, pharmacokinetic properties, and toxicity of identified candidates.

Main Methods:

• Machine learning algorithms (Random Forest, Extra Trees, Nu-Support Vector) were employed to build a voting classifier.
• A virtual screening of 10 million molecules was performed, followed by filtering for toxicity.
• Molecular docking, MM/PBSA, DFT calculations, and LD50 estimations were used to assess candidate properties.

Main Results:

• Two compounds, L1 and L2, exhibited strong binding affinities to MtbMetRS (-12.74 kcal/mol for L1, -11.83 kcal/mol for L2).
• These candidates demonstrated favorable pharmacokinetic profiles and predicted stability.
• Computational analyses indicated a safer toxicity profile for L1 and L2.

Conclusions:

• L1 and L2 are identified as promising potential inhibitors of MtbMetRS.
• These compounds represent viable starting points for developing new anti-tuberculosis drugs.
• Further in vitro and in vivo studies are required to validate their efficacy and safety.