Machine Learning-Assisted High-Throughput Screening for Anti-MRSA Compounds.

Area of Science:

• Computational chemistry
• Drug discovery
• Machine learning applications

Background:

• Antimicrobial resistance (AMR) poses a significant global health challenge, necessitating novel therapeutic agents.
• Traditional compound screening methods are time-consuming and expensive.
• Computational approaches offer a promising strategy to expedite the identification of new antimicrobial compounds.

Purpose of the Study:

• To develop and validate a machine learning (ML) model for the in silico screening of low molecular weight compounds.
• To enhance the efficiency and reduce the cost of identifying potential antimicrobial agents.

Main Methods:

• Utilized results from a high-throughput Caenorhabditis elegans methicillin-resistant Staphylococcus aureus (MRSA) liquid infection assay.
• Developed ML models for compound prioritization and quality control.
• Applied the model to a large validation set of chemical compounds.

Main Results:

• The compound prioritization model demonstrated strong performance with an AUC of 0.795, 81% sensitivity, and 70% specificity.
• The ML model identified 81% of active compounds within only 30.6% of the tested compounds, increasing the hit rate by 2.67-fold.
• Retraining the model identified 42 out of 45 (93%) previously missed active compounds.

Conclusions:

• The developed ML approach significantly improves the efficiency of high-throughput screening (HTS).
• This method reduces the number of compounds requiring physical screening and identifies potentially overlooked active molecules.
• The ML model makes HTS more accessible, lowering barriers to entry for antimicrobial drug discovery.