PPB-Affinity: Protein-Protein Binding Affinity dataset for AI-based protein drug discovery

Area of Science:

• Biochemistry
• Computational Biology
• Drug Discovery

Background:

• Protein-protein binding (PPB) affinity prediction is crucial for large-molecule drug discovery.
• Deep learning models exist for predicting affinity changes due to mutations, but not for predicting absolute PPB affinity.
• A lack of publicly available datasets hinders research in predicting PPB affinity.

Purpose of the Study:

• To introduce a comprehensive, large-scale, publicly available dataset for protein-protein binding affinity (PPB-Affinity).
• To address the scarcity of data for predicting PPB affinity.
• To facilitate advancements in large-molecule drug discovery and screening.

Main Methods:

• Compilation of a novel dataset (PPB-Affinity) containing protein-protein complex crystal structures, mutation data, and binding affinities.
• Inclusion of detailed information on receptor and ligand protein chains.
• Development of a deep learning benchmark model utilizing the PPB-Affinity dataset.

Main Results:

• Creation of the largest publicly available dataset specifically for protein-protein binding affinity.
• The PPB-Affinity dataset includes structural and affinity data for protein complexes.
• A deep learning benchmark model was established for PPB affinity prediction.

Conclusions:

• The PPB-Affinity dataset represents a significant resource for the research community.
• This dataset is expected to accelerate the screening of potential large-molecule drugs.
• The benchmark model provides a foundation for future research in PPB affinity prediction.