Machine Learning Models Identify Inhibitors of New Delhi Metallo-β-lactamase.

Area of Science:

• Antimicrobial resistance
• Drug discovery
• Computational chemistry

Background:

• Metallo-β-lactamases (MBLs), particularly NDM-1, are a growing global threat to the effectiveness of crucial β-lactam antibiotics.
• The clinical pipeline lacks approved MBL inhibitors despite extensive research using methods like high-throughput screening (HTS) and fragment-based drug discovery (FBDD).

Purpose of the Study:

• To develop and apply a machine learning (ML) based prediction tool for identifying novel metallo-β-lactamase inhibitors.
• To validate the efficacy of a computationally identified inhibitor against NDM-1 producing bacteria.

Main Methods:

• A machine learning model was trained on HTS data and published inhibition data.
• Virtual screening of the NIH Genesis library was performed using the ML tool, followed by quantitative HTS (qHTS).
• Mechanism of inhibition was investigated using biophysical techniques and molecular docking.

Main Results:

• A novel MBL inhibitor, compound 72922413, was identified through ML-driven virtual screening and qHTS.
• Compound 72922413 demonstrated the ability to reduce meropenem minimum inhibitory concentrations (MICs) in clinical isolates of *E. coli* and *K. pneumoniae* expressing NDM-1.
• The study elucidated the inhibition mechanism of this novel pyrido[1,2-a]pyrimidin-4-one scaffold.

Conclusions:

• Machine learning combined with HTS is an effective strategy for discovering novel antibiotic resistance inhibitors.
• Compound 72922413 represents a promising lead for developing new therapies against NDM-1-mediated carbapenem resistance.
• Further development of this inhibitor could help combat the rising threat of multidrug-resistant bacteria.