Protein-Protein Interface Topology as a Predictor of Secondary Structure and Molecular Function Using Convolutional...

Area of Science:

• Computational biology
• Structural biology
• Bioinformatics

Background:

• Protein-protein interactions are fundamental to cellular processes.
• The shape and topology of protein-protein interfaces encode information about binding specificity and function.
• Predicting these properties from interface structure is a key challenge in bioinformatics.

Purpose of the Study:

• To investigate if the 3D shape of protein-protein interfaces contains sufficient information to predict associated secondary structure motifs and molecular functions.
• To develop and apply deep learning methods for analyzing protein interface geometry.

Main Methods:

• Utilized convolutional deep learning on 3D voxel representations of protein-protein interfaces.
• Employed a novel two-stage network processing interfaces at two resolutions to balance performance and computational cost.
• Colored voxel representations by burial depth to provide structural context.

Main Results:

• The deep learning network accurately predicted secondary structure motifs at interfaces.
• Certain classes of molecular function were also well predicted based on interface shape.
• Identified specific interface patterns crucial for recognizing distinct protein classes.

Conclusions:

• The global shape and local topological features of protein-protein interfaces are rich sources of information about their function.
• Deep learning on 3D structural data is a powerful approach for deciphering protein interaction determinants.
• Interface geometry directly correlates with higher-level functional and structural information.