Combining Docking Pose Rank and Structure with Deep Learning Improves Protein–Ligand Binding Mode Prediction over a Baseline Docking Approach

Abstract

We present a simple, modular graph-based convolutional neural network that takes structural information from protein-ligand complexes as input to generate models for activity and binding mode prediction. Complex structures are generated by a standard docking procedure and fed into a dual-graph architecture that includes separate sub-networks for the ligand bonded topology and the ligand-protein contact map. This network division allows contributions from ligand identity to be distinguished from effects of protein-ligand interactions on classification. We show, in agreement with recent literature, that dataset bias drives many of the promising results on virtual screening that have previously been reported. However, we also show that our neural network is capable of learning from protein structural information when, as in the case of binding mode prediction, an unbiased dataset is constructed. We develop a deep learning model for binding mode prediction that uses docking ranking as input in combination with docking structures. This strategy mirrors past consensus models and outperforms the baseline docking program in a variety of tests, including on cross-docking datasets that mimic real-world docking use cases. Furthermore, the magnitudes of network predictions serve as reliable measures of model confidence.

Computational Modeling as a Tool to Investigate PPI: From Drug Design to Tissue Engineering

“…How these decoy sets are generated influences the corresponding scoring functions, often resulting in biases. Recent efforts aim at detecting and overcoming such biases ( Morrone et al, 2020 ). Scoring functions that work for protein-protein or protein-small molecule systems are not always transferable to protein-peptide systems, resulting in the development of several specific peptide-protein scoring functions ( Raveh et al, 2011 ; Kurcinski et al, 2015 ; Spiliotopoulos et al, 2016 ; Tao et al, 2020 ).…”

3D Convolutional Neural Networks and a CrossDocked Dataset for Structure-Based Drug Design

“…cheminformatic) models 19. A more careful process for generating putative ligand poses, such as template-based docking, 49 may yield better results.Alternatively, a different input representation, model architecture, or training regime might force a model to only predict using protein-ligand interaction information,50 although it is not clear this necessarily results in a more generalizable model as ligand-only information is embedded in protein-ligand interaction information.…”

A Deep-Learning Approach toward Rational Molecular Docking Protocol Selection

“…Finally, while we firmly believe that future-generation docking protocols will more tightly incorporate machine-learning elements into their pipelines [18,19] (e.g., by the design of more efficient search algorithms or scoring functions [55,56]), we think that the approach proposed in this paper represents a novel research direction that will drive structure-based drug design researchers towards more rational existing docking protocol choices. Hence, with the intent of improving research reproducibility and lowering accessibility barriers, we have open-sourced all evaluation and deployment code as well as trained models related to this work.…”