DeepRank-GNN: a graph neural network framework to learn patterns in protein-protein interfaces

Abstract

Motivation: Gaining structural insights into the protein-protein interactome is essential to understand biological phenomena and extract knowledge for rational drug design or protein engineering. We have previously developed DeepRank, a deep-learning framework to facilitate pattern learning from protein-protein interfaces using convolutional neural network (CNN) approaches. However, CNN is not rotation invariant and data augmentation is required to desensitize the network to the input data orientation which dramatically impairs the computation performance. Representing protein-protein complexes as atomic- or residue-scale rotation invariant graphs instead enables using graph neural networks (GNN) approaches, bypassing those limitations.

Results: We have developed DeepRank-GNN, a framework that converts protein-protein interfaces from PDB 3D coordinates files into graphs that are further provided to a pre-defined or user-defined GNN architecture to learn problem-specific interaction patterns. DeepRank-GNN is designed to be highly modularizable, easily customized and is wrapped into a user-friendly python3 package. Here, we showcase DeepRank-GNN's performance on two applications using a dedicated graph interaction neural network: (i) the scoring of docking poses and (ii) the discriminating of biological and crystal interfaces. In addition to the highly competitive performance obtained in those tasks as compared to state-of-the-art methods, we show a significant improvement in speed and storage requirement using DeepRank-GNN as compared to DeepRank.

Availability and implementation: DeepRank-GNN is freely available from https://github.com/DeepRank/DeepRank-GNN.

Supplementary information: Supplementary data are available at Bioinformatics online.

Fig. 1.

Overview of the DeepRank-GNN framework. (A) DeepRank-GNN identifies interface residues and converts them into an interface graph. Internal edges are defined between residues from the same chain having heavy atoms within a 3 Å distance cutoff from each other, while external edges are defined between residues from different chains having heavy atoms within the 8.5 Å cutoff. (B) Example of GNN architecture (GINet). The graph representation of a PPI is split into two sub-graphs, i.e. the internal graph connecting atoms from the same protein and the external graph connecting atoms from distinct proteins. The two sub-graphs are sequentially passed to two consecutive convolution/activation/pooling layers. The two final graph representations are flattened using the mean value of each feature and merged before applying two fully connected layers. GCL, graph convolution layer; FCC, fully connected layer

Fig. 2.

Comparison of DeepRank-GNN with HADDOCK scoring function on the BM5 set. Average ROC curves obtained with the models retained for each DeepRank-GNN fold, for the model trained on the full training set and HADDOCK score. A true positive case corresponds to a complex with f_nat > 0.3 correctly predicted. The number of true positive rate values is averaged over the number of complexes in the test dataset. The dashed line represents a random classifier