CrypTothML

CrypTothML integrates mixed-solvent molecular dynamics simulations and machine learning to predict cryptic sites in proteins for drug discovery applications.

Key Features:

• MSMD-based hotspot identification: Identifies hotspots using mixed-solvent molecular dynamics simulations with six chemically diverse probes (benzene, dimethyl-ether, phenol, methyl-imidazole, acetonitrile, and ethylene glycol).
• Machine learning ranking: Ranks hotspots based on their likelihood of being cryptic sites using hotspot-derived and protein-specific features.
• Performance evaluation: Achieves an AUC-ROC of 0.88 on a curated dataset and ranks cryptic sites as the top prediction more frequently than existing methods.
• Cryptic site detection: Successfully identifies cryptic sites missed by other machine learning-based methods.

Scientific Applications:

• Allosteric drug discovery: Identifies transient binding locations that emerge through protein conformational changes upon ligand binding for allosteric modulator development.
• Cryptic site prediction: Predicts cryptic sites without requiring experimentally determined holo and apo states.

Methodology:

CrypTothML first identifies hotspots through mixed-solvent molecular dynamics simulations using six chemically diverse probes. A machine learning model then ranks these hotspots based on their likelihood of being cryptic sites, incorporating both hotspot-derived and protein-specific features.