AI3: 20,000 Protein-Ligand Complexes with Simulated Binding Affinities

The AI3 Protein-Ligand Binding Affinity Dataset was created by teams from IIIT Hyderabad, Intel, AWS, and Insilico Medicine. It contains physics-based calculations on approximately 20,000 protein-ligand complexes, including molecular dynamics snapshots and binding affinities calculated via the MM-PBSA method. The dataset provides 3D coordinates and CSV files with energy components like electrostatic and van der Waals interactions.

Use Cases

• Train ML scoring functions to predict binding affinities based on molecular dynamics snapshots and energy components.
• Build classification models for predicting correct ligand binding poses using the 3D coordinate data.
• Identify cryptic binding pockets in proteins by analyzing the dynamic simulation data.
• Optimize binding features by exploiting individual energy components, which are not typically available in experimental data.

Strengths

• Contains data for approximately 20,000 protein-ligand complexes.
• Includes dynamic molecular dynamics simulation snapshots, which are noted as lacking in existing training datasets.
• Provides individual energy components like electrostatic and van der Waals interactions, beyond just total binding affinity.
• Data generated from consistent computational protocols, reducing bias from varied experimental methods.

Limitations

• Row count and dataset size are unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• Last update date is unknown; freshness unverified.

Provenance

Source

International Institute of Information Technology Hyderabad, Intel, AWS, and Insilico Medicine

Collection Method

Physics-based molecular dynamics simulations and MM-PBSA free energy calculations.

Time Range

null

Freshness

null

Geography

null