Comparison of Target Features for Predicting Drug-Target Interactions by Deep Neural Network Based on Large-Scale Drug-Induced Transcriptome Data.

Hanbi Lee,Wankyu Kim

doi:10.3390/pharmaceutics11080377

Hanbi Lee, Wankyu Kim

Open Access

https://doi.org/10.3390/pharmaceutics11080377

Copy DOI

Journal: Pharmaceutics	Publication Date: Aug 2, 2019
Citations: 27	License type: CC BY 4.0

Affiliation: Ewha Womans University

Abstract

Uncovering drug-target interactions (DTIs) is pivotal to understand drug mode-of-action (MoA), avoid adverse drug reaction (ADR), and seek opportunities for drug repositioning (DR). For decades, in silico predictions for DTIs have largely depended on structural information of both targets and compounds, e.g., docking or ligand-based virtual screening. Recently, the application of deep neural network (DNN) is opening a new path to uncover novel DTIs for thousands of targets. One important question is which features for targets are most relevant to DTI prediction. As an early attempt to answer this question, we objectively compared three canonical target features extracted from: (i) the expression profiles by gene knockdown (GEPs); (ii) the protein–protein interaction network (PPI network); and (iii) the pathway membership (PM) of a target gene. For drug features, the large-scale drug-induced transcriptome dataset, or the Library of Integrated Network-based Cellular Signatures (LINCS) L1000 dataset was used. All these features are closely related to protein function or drug MoA, of which utility is only sparsely investigated. In particular, few studies have compared the three types of target features in DNN-based DTI prediction under the same evaluation scheme. Among the three target features, the PM and the PPI network show similar performances superior to GEPs. DNN models based on both features consistently outperformed other machine learning methods such as naïve Bayes, random forest, or logistic regression.

Highlights

In silico prediction of drug–target interaction (DTI) is becoming more data-driven than conventional modeling-based approaches such as docking or molecular dynamic simulation
drug-induced expression profiles (DEPs) were set common to all the deep neural network (DNN) models or other machine learning techniques, so that different target features are compared under the same condition
DTI prediction is critical for efficient hit discovery, target deconvolution after phenotypic screening, and identifying novel indication in drug repositioning

Summary

Introduction

In silico prediction of drug–target interaction (DTI) is becoming more data-driven than conventional modeling-based approaches such as docking or molecular dynamic simulation. The vast majority of drugs and compounds are expected to interact with multiple targets, i.e., polypharmacology [2]. While millions of DTIs have been identified, and increasingly keep being revealed, it is still costly and time-consuming to validate DTIs experimentally even by high throughput screening (HTS) [3]. It is most likely that there still exist unknown DTIs for both approved drugs and clinical candidate compounds. Such hidden DTIs could critically impact the drug development process including unexpected clinical outcome, or may broaden their indications through drug repositioning

Methods

Results

Conclusion