Drug-target interaction prediction using semi-bipartite graph model and deep learning

Hafez Eslami Manoochehri,Mehrdad Nourani

doi:10.1186/s12859-020-3518-6

Abstract

BackgroundIdentifying drug-target interaction is a key element in drug discovery. In silico prediction of drug-target interaction can speed up the process of identifying unknown interactions between drugs and target proteins. In recent studies, handcrafted features, similarity metrics and machine learning methods have been proposed for predicting drug-target interactions. However, these methods cannot fully learn the underlying relations between drugs and targets. In this paper, we propose anew framework for drug-target interaction prediction that learns latent features from drug-target interaction network.ResultsWe present a framework to utilize the network topology and identify interacting and non-interacting drug-target pairs. We model the problem as a semi-bipartite graph in which we are able to use drug-drug and protein-protein similarity in a drug-protein network. We have then used a graph labeling method for vertex ordering in our graph embedding process. Finally, we employed deep neural network to learn the complex pattern of interacting pairs from embedded graphs. We show our approach is able to learn sophisticated drug-target topological features and outperforms other state-of-the-art approaches.ConclusionsThe proposed learning model on semi-bipartite graph model, can integrate drug-drug and protein-protein similarities which are semantically different than drug-protein information in a drug-target interaction network. We show our model can determine interaction likelihood for each drug-target pair and outperform other heuristics.

Highlights

Identifying drug-target interaction is a key element in drug discovery
Inspired by link prediction methods for complex graphs, in this paper we propose a supervised learning heuristic for drug-target interaction prediction that unlike traditional methods that rely on hand-engineered graph features, it learns the network topology by itself
To capture the drug-drug and target-target similarities, we formulate the Drug-Target Interactions (DTI) network as an un-directed semi-bipartite graph G =< D, T, E, F, H >, where D and T are set of drug and target nodes respectively, E ⊂ D×T is the set of edges between D and T, i.e. E = {(di, tj)|di ∈ D, tj ∈ T}, F ⊂ D × D is the set of edges between the nodes in D, i.e. F = {(di, dj)|di, dj ∈ D} and H ⊂ T ×T is the set of edges between the nodes in T, i.e. H = {(ti, tj)|ti, tj ∈ T}

Summary

Introduction

Identifying drug-target interaction is a key element in drug discovery. In silico prediction of drug-target interaction can speed up the process of identifying unknown interactions between drugs and target proteins. In recent studies, handcrafted features, similarity metrics and machine learning methods have been proposed for predicting drug-target interactions. These methods cannot fully learn the underlying relations between drugs and targets. Most of traditional approaches for predicting DTI, either for drug discovery or repositioning (reusing already available drugs for new targets) are ligand-based approaches. Ligand’s and docking methods run simulations to estimate the likelihood that it will interact with a certain drug based on their binding affinity and strength [3, 4] These approaches often lead to poor prediction results when a target has only a small number of known binding ligands. The performance of docking-based approaches is limited to availability of 3D structures of target proteins and can be quite poor

Methods

Results

Conclusion