DTi2Vec: Drug-target interaction prediction using network embedding and ensemble learning

Abstract

Drug-target interaction (DTI) prediction is a crucial step in drug discovery and repositioning as it reduces experimental validation costs if done right. Thus, developing in-silico methods to predict potential DTI has become a competitive research niche, with one of its main focuses being improving the prediction accuracy. Using machine learning (ML) models for this task, specifically network-based approaches, is effective and has shown great advantages over the other computational methods. However, ML model development involves upstream hand-crafted feature extraction and other processes that impact prediction accuracy. Thus, network-based representation learning techniques that provide automated feature extraction combined with traditional ML classifiers dealing with downstream link prediction tasks may be better-suited paradigms. Here, we present such a method, DTi2Vec, which identifies DTIs using network representation learning and ensemble learning techniques. DTi2Vec constructs the heterogeneous network, and then it automatically generates features for each drug and target using the nodes embedding technique. DTi2Vec demonstrated its ability in drug-target link prediction compared to several state-of-the-art network-based methods, using four benchmark datasets and large-scale data compiled from DrugBank. DTi2Vec showed a statistically significant increase in the prediction performances in terms of AUPR. We verified the "novel" predicted DTIs using several databases and scientific literature. DTi2Vec is a simple yet effective method that provides high DTI prediction performance while being scalable and efficient in computation, translating into a powerful drug repositioning tool.

Keywords: Cheminformatics; Drug repositioning; Drug–target interaction; Ensemble learning; Heterogeneous network; Link prediction; Network embedding; Random walk; Representation learning.

Fig. 1

An illustration of the underlying DTI link prediction paradigm

Fig. 2

DTi2Vec Method Flowchart, (1) Filter the TTsim and DDsim graphs, (2) Construct a full DTI network by augmenting the three graphs, (3) Apply the three-step node2vec framework on the full DTI network, (4) Generate edge representation for each drug-target pair, (5) Feed the feature vector (FV) into ensemble boosting classifier to output the class labels

Fig. 3

Visualization of similarity matrices of the NR dataset, a drug-drug similarity matrix, b target–target similarity matrix

Fig. 4

A depiction of the datasets’ sparsity ratios and the model performances in terms of AUPR applied in AdaBoost and XGBoost classifiers on Hadamard FVs

Fig. 5

Comparing the prediction performance of DTi2Vec and state-of-the-art methods in random CV setting (in terms of AUPR with standard errors are shown) using the Yamanishi_08 and FDA_DrugBank datasets

Fig. 6

Comparing the prediction performance of DTi2Vec and other state-of-the-art methods when using new drug settings (in terms of AUPR using the Yamanishi_08 and FDA_DrugBank datasets, and the average AUPR across all datasets)