AutoDTI++: deep unsupervised learning for DTI prediction by autoencoders

Abstract

Background Drug–target interaction (DTI) plays a vital role in drug discovery. Identifying drug–target interactions related to wet-lab experiments are costly, laborious, and time-consuming. Therefore, computational methods to predict drug–target interactions are an essential task in the drug discovery process. Meanwhile, computational methods can reduce search space by proposing potential drugs already validated on wet-lab experiments. Recently, deep learning-based methods in drug-target interaction prediction have gotten more attention. Traditionally, DTI prediction methods' performance heavily depends on additional information, such as protein sequence and molecular structure of the drug, as well as deep supervised learning. Results This paper proposes a method based on deep unsupervised learning for drug-target interaction prediction called AutoDTI++. The proposed method includes three steps. The first step is to pre-process the interaction matrix. Since the interaction matrix is sparse, we solved the sparsity of the interaction matrix with drug fingerprints. Then, in the second step, the AutoDTI approach is introduced. In the third step, we post-preprocess the output of the AutoDTI model. Conclusions Experimental results have shown that we were able to improve the prediction performance. To this end, the proposed method has been compared to other algorithms using the same reference datasets. The proposed method indicates that the experimental results of running five repetitions of tenfold cross-validation on golden standard datasets (Nuclear Receptors, GPCRs, Ion channels, and Enzymes) achieve good performance with high accuracy.

A brief review of protein–ligand interaction prediction

“…Supervised learning requires large-scale high-quality labeled datasets. In the case of an insufficient quantity of labeled PLI datasets, research works apply unsupervised learning, semi-supervised learning, or self-supervised learning to predict the PLIs [64] , [65] , [66] . In particular, unsupervised pretrained models on large text corpora have shown remarkable performance on various natural language processing tasks.…”

Proteome‐Scale Drug‐Target Interaction Predictions: Approaches and Applications

“…In their model, unseen proteins have the largest impact on predictive performance. AutoDTI++ also reports performance of ligand-and protein-exclusion studies (Sajadi, Chahooki, Gharaghani, & Abbasi, 2021), but notice a more significant loss in predictive power in the ligand exclusion test. AutoDTI++, however, differs from most other DTI prediction tools reviewed in this study in its lack of protein features and by framing DTI predictions as an unsupervised imputation task.…”

Multi-Objective Drug Design Based on Graph-Fragment Molecular Representation and Deep Evolutionary Learning

“…Recently, machine learning approaches have been exploited to predict drug-target interaction (DTI) or drug-target binding affinity (DTBA) through learning on heterogeneous biological data of known interactions to understand the mechanism of drug actions. For example, AutoDTI++ ( Sajadi et al, 2021 ) uses a denoising autoencoder that reconstructs the drug-target interaction matrix by adopting denoising empirical loss, which emphasizes interaction prediction while discarding the loss of missing values. The model input is composed by multiplying the drug-target interaction matrix by the fingerprint-drug matrix for additional information on drug fingerprints.…”