EquiPPIS

EquiPPIS predicts protein–protein interaction (PPI) sites using E(3) equivariant graph neural networks to provide residue-level interfacial predictions from experimental and AlphaFold2-predicted structures.

Key Features:

• E(3) equivariance: Employs symmetry-aware graph convolutions that transform equivariantly under translations, rotations, and reflections in three-dimensional space.
• Robustness and accuracy: Demonstrates improved predictive accuracy compared with state-of-the-art methods on the same experimental inputs and achieves higher accuracy on AlphaFold2-predicted structural models.
• Scalability: Supports large-scale prediction of PPI sites across multiple structures and datasets.

Scientific Applications:

• Interfacial residue prediction: Predicts residue-level interfaces involved in protein-protein interactions for structural characterization.
• Mechanistic studies: Aids interpretation of molecular mechanisms by identifying interaction sites that mediate biological processes.
• Drug discovery and therapeutic design: Supports identification of PPI targets and design of molecules to modulate protein-protein interactions.

Methodology:

Applies graph neural networks with integrated E(3) equivariance implemented as symmetry-aware graph convolutions that preserve equivariant behavior under translation, rotation, and reflection to learn spatially consistent residue representations from experimental and AlphaFold2-predicted structures.