Drug-Target Interaction prediction using Multi-Graph Regularized Deep Matrix Factorization

Abstract

Drug discovery is an important field in the pharmaceutical industry with one of its crucial chemogenomic process being drug-target interaction prediction. This interaction determination is expensive and laborious, which brings the need for alternative computational approaches which could help reduce the search space for biological experiments. This paper proposes a novel framework for drug-target interaction (DTI) prediction: Multi-Graph Regularized Deep Matrix Factorization (MGRDMF). The proposed method, motivated by the success of deep learning, finds a low-rank solution which is structured by the proximities of drugs and targets (drug similarities and target similarities) using deep matrix factorization. Deep matrix factorization is capable of learning deep representations of drugs and targets for interaction prediction. It is an established fact that drug and target similarities incorporation preserves the local geometries of the data in original space and learns the data manifold better. However, there is no literature on which the type of similarity matrix (apart from the standard biological chemical structure similarity for drugs and genomic sequence similarity for targets) could best help in DTI prediction. Therefore, we attempt to take into account various types of similarities between drugs/targets as multiple graph Laplacian regularization terms which take into account the neighborhood information between drugs/targets. This is the first work which has leveraged multiple similarity/neighborhood information into the deep learning framework for drug-target interaction prediction. The cross-validation results on four benchmark data sets validate the efficacy of the proposed algorithm by outperforming shallow state-of-the-art computational methods on the grounds of AUPR and AUC.September 18, 2019 1/17 interaction not only assists drug discovery but also affects other fields such as drug repositioning, drug resistance and side-effect prediction [10]. As an example, Drug repositioning [11,12] (using an existing drug for new indications) can grant polypharmacology (multi-target effect) to a drug. Gleevec (imatinib mesylate) is one of the many such examples which was successfully repositioned. Earlier, it was known to interact only with the Bcr-Abl fusion gene which is indicative of leukemia. However, later discoveries showing that it also interacts with PDGF and KIT, repositioned it for the treatment of gastrointestinal stromal tumors [13,14]. The methods available for prediction of DTI can be divided into the following three broad categories: Ligand-based approaches, Docking based approaches, and Chemogenomic approaches. Ligand-based approaches predict interactions by deploying the similarity between the ligands of target proteins [15]. The idea is that molecules with similar structure/property would bind similar proteins [16]. But, the reliability of results might get compromised due to limited information about known ligands per protein. Docking-based approaches use the three-dimensional structure of both ...

DeepVir - Graphical Deep Matrix Factorization for In Silico Antiviral Repositioning: Application to COVID-19

“…In this subsection, we study the choice of technique to solve the GRDMF problem (10). In the past work [50,49], the proposed formulation has been solved using ADMM (alternating direction method of multipliers) [8,37]. We introduce here a novel resolution strategy named HyPALM.…”

“…A prior study on graph regularised deep matrix factorization [49] was based on the alternating direction method of multipliers (ADMM) approach [76]. A major issue with ADMM is that the convergence guarantees are rather mild in such challenging non convex scenario.…”

Using Single Protein/Ligand Binding Models to Predict Active Ligands for Unseen Proteins

“…Performance of these approaches at generalizing to proteins or ligands without experimental data is poor (21-32). Moreover, methods for DTI prediction have largely been tested using a pair splitting methodology, where the dataset of (protein, ligand) pairs is split into train, validation, and test sets (16,19,(24)(25)(26)(32)(33)(34)(35)(36)(37)(38)(39)(40)(41)(42). This approach does not determine whether models can generalize to unseen protein targets, since the test set contains protein targets for which many active ligands are seen during training.…”