Protein Docking Model Evaluation by Graph Neural Networks

Abstract

Physical interactions of proteins play key roles in many important cellular processes. Therefore, it is crucial to determine the structure of protein complexes to understand molecular mechanisms of interactions. To complement experimental approaches, which usually take a considerable amount of time and resources, various computational methods have been developed to predict the structures of protein complexes. In computational modeling, one of the challenges is to identify near-native structures from a large pool of generated models. Here, we developed a deep learning-based approach named Graph Neural Network-based DOcking decoy eValuation scorE (GNN-DOVE). To evaluate a protein docking model, GNN-DOVE extracts the interface area and represents it as a graph. The chemical properties of atoms and the inter-atom distances are used as features of nodes and edges in the graph. GNN-DOVE was trained and validated on docking models in the Dockground database. GNN-DOVE performed better than existing methods including DOVE, which is our previous development that uses convolutional neural network on voxelized structure models.

Benchmarking of structure refinement methods for protein complex models

“…The common drive of the current methods has been classical MD or coarse-grained dynamics simulation. As we observe a rapid development and improvement of deep-learning-based structure modeling methods and quality assessment methods in CASP 36,37,40,41 and CAPRI, 11,42 we expect that new structure refinement strategies using deep learning will appear in the near future.…”

DLA-Ranker: Evaluating protein docking conformations with many locally oriented cubes

“…To limit this effect, (Renaud et al ., 2021) performed rotational data augmentation. Some other works have proposed alternative representations such as graphs (Wang et al ., 2021; Cao and Shen, 2020) or point clouds (Eismann et al ., 2021). By contrast to our local-based approach, they adopt a global perspective by assessing the quality of the interface (Wang et al ., 2021) or even the complex (Cao and Shen, 2020; Eismann et al ., 2021) as a whole.…”

“…Some other works have proposed alternative representations such as graphs (Wang et al ., 2021; Cao and Shen, 2020) or point clouds (Eismann et al ., 2021). By contrast to our local-based approach, they adopt a global perspective by assessing the quality of the interface (Wang et al ., 2021) or even the complex (Cao and Shen, 2020; Eismann et al ., 2021) as a whole. Representing the input conformation as a graph (Cao and Shen, 2020; Wang et al ., 2021) renders the prediction rotationally invariant, but the information of the relative orientations of the atoms in the structure is lost.…”

“…By contrast to our local-based approach, they adopt a global perspective by assessing the quality of the interface (Wang et al ., 2021) or even the complex (Cao and Shen, 2020; Eismann et al ., 2021) as a whole. Representing the input conformation as a graph (Cao and Shen, 2020; Wang et al ., 2021) renders the prediction rotationally invariant, but the information of the relative orientations of the atoms in the structure is lost. In the point cloud-based method (Eismann et al ., 2021), the authors eliminate the need for rotational data augmentation by using tensor field network layers where the convolutional filters are decomposed into series of spherical harmonics.…”

Directed Weight Neural Networks for Protein Structure Representation Learning

“…Therefore, processing such 3D structures is the key for protein function analysis. While we have witnessed remarkable progress in protein structure predictions (Rohl et al, 2004;Källberg et al, 2012;Baek et al, 2021;Jumper et al, 2021), another thread of tasks with protein 3D structures as input starts to draw a great interest, such as function prediction (Hermosilla et al, 2020;Gligorijević et al, 2021), decoy ranking (Lundström et al, 2001;Kwon et al, 2021;Wang et al, 2021), protein docking (Duhovny et al, 2002;Shulman-Peleg et al, 2004;Gainza et al, 2020;Sverrisson et al, 2021), and driver mutation identification (Lefèvre et al, 1997;Antikainen & Martin, 2005;Li et al, 2020;Jankauskaitė et al, 2019).…”