DockNet: high-throughput protein-protein interface contact prediction.

Area of Science:

• Computational biology
• Bioinformatics
• Drug discovery

Background:

• Over 300,000 human protein-protein interaction (PPI) pairs are known, presenting a key target for drug design.
• Predicting PPI sites is crucial but challenging, with traditional docking simulations being computationally expensive and often failing to account for protein flexibility.
• Existing methods have limitations in handling protein flexibility, leading to suboptimal prediction accuracy.

Purpose of the Study:

• To develop an efficient and accurate method for predicting protein-protein interaction sites.
• To address the limitations of current methods by incorporating full protein structure and flexibility.
• To provide a tool that can be utilized even when accurate unbound protein structures are unavailable.

Main Methods:

• Proposed DockNet, a Siamese graph-based neural network.
• Incorporated entire protein structures, allowing for unlimited protein flexibility during interaction.
• Modeled predictions at the residue level using diverse node features (residue type, surface accessibility, depth, secondary structure, pharmacophore, torsional angles).

Main Results:

• DockNet achieved an Area Under the Curve (AUC) of 0.84 on an independent test set (DB5).
• The method demonstrates comparable performance to state-of-the-art techniques.
• DockNet is versatile and applicable to various protein structures and scenarios lacking accurate unbound structures.

Conclusions:

• DockNet offers an efficient and flexible approach to predicting protein-protein interaction sites.
• The method advances the field of computational drug design by improving PPI prediction accuracy.
• DockNet is accessible via GitHub and a webserver for broader research application.