TACTICS

TACTICS identifies cryptic druggable pockets in proteins from molecular dynamics trajectories by combining k-means clustering, a random forest model, and fragment docking to prioritize potentially bindable residues.

Key Features:

• Input Data Utilization: Processes ensembles of molecular conformations from molecular dynamics simulations to capture conformational flexibility and transient binding pockets.
• Conformation Selection via k-means Clustering: Applies k-means clustering to distill a representative subset of conformations reflecting the protein's conformational heterogeneity.
• Machine Learning (Random Forest): Uses a random forest model to predict potentially bindable residues for each selected conformation, considering protein motion and geometric properties.
• Scoring with Fragment Docking: Evaluates and scores identified residues/sites using fragment docking to assess their potential for stable small-molecule interactions.
• Cryptic Pocket Identification: Detects binding sites not evident in static, experimentally determined protein structures.

Scientific Applications:

• SARS-CoV-2 main protease: Applied to the SARS-CoV-2 main protease to recapitulate known small-molecule binding sites and predict novel cryptic pockets.
• SARS-CoV-2 methyltransferase: Applied to the SARS-CoV-2 methyltransferase with identification of known and previously unobserved pockets.
• Yersinia pestis aryl carrier protein: Applied to the Yersinia pestis aryl carrier protein to recover known binding sites and predict additional cryptic pockets.

Methodology:

Processes molecular dynamics ensembles, applies k-means clustering to select representative conformations, uses a random forest model to predict bindable residues based on motion and geometric features, and evaluates candidates by fragment docking.