CAVIAR

CAVIAR identifies and characterizes protein binding cavities and subcavities from PDB/mmCIF structures and molecular dynamics trajectories to generate binding-site descriptors for ligandability assessment, hit identification, and affinity prediction.

Key Features:

• Automatic Cavity Detection: Detects cavities from Protein Data Bank (PDB) and mmCIF files and from molecular dynamics trajectory frames.
• Subcavity Segmentation: Applies a subcavity segmentation algorithm that decomposes cavities into empirically relevant subpockets aligned with medicinal chemistry definitions.
• Descriptor Generation: Computes binding-site descriptors intended for downstream computational analyses and machine learning.
• Machine Learning Integration: Produces descriptors suitable for machine learning applications including hit identification, protein engineering, and prediction of experimental binding affinities.
• Ligand Independence: Assigns subcavities and generates descriptors without requiring bound ligand information.

Scientific Applications:

• Medicinal Chemistry: Identification and characterization of subpockets to support rational drug design and ligandability assessment.
• Protein Engineering: Structural analysis of binding sites to guide modifications for altered functionality or stability.
• Binding Affinity Prediction: Descriptor-based analyses that correlate with experimental metrics, including the relationship between subcavities filled and high-affinity (e.g., nanomolar) interactions.
• Hit Identification: Use of binding-site descriptors in computational workflows to prioritize compounds or fragments for further study.

Methodology:

Processes PDB and mmCIF structures and molecular dynamics frames; performs cavity detection and subcavity segmentation; computes binding-site descriptors for machine learning and affinity-related analyses; assigns subcavities without ligand input.