DTI-Voodoo: machine learning over interaction networks and ontology-based background knowledge predicts drug–target interactions

Abstract

Motivation In silico drug–target interaction (DTI) prediction is important for drug discovery and drug repurposing. Approaches to predict DTIs can proceed indirectly, top-down, using phenotypic effects of drugs to identify potential drug targets, or they can be direct, bottom-up and use molecular information to directly predict binding affinities. Both approaches can be combined with information about interaction networks. Results We developed DTI-Voodoo as a computational method that combines molecular features and ontology-encoded phenotypic effects of drugs with protein–protein interaction networks, and uses a graph convolutional neural network to predict DTIs. We demonstrate that drug effect features can exploit information in the interaction network whereas molecular features do not. DTI-Voodoo is designed to predict candidate drugs for a given protein; we use this formulation to show that common DTI datasets contain intrinsic biases with major effects on performance evaluation and comparison of DTI prediction methods. Using a modified evaluation scheme, we demonstrate that DTI-Voodoo improves significantly over state of the art DTI prediction methods. Availability DTI-Voodoo source code and data necessary to reproduce results are freely available at https://github.com/THinnerichs/DTI-VOODOO. Supplementary information Supplementary data are available at Bioinformatics online.

Artificial intelligence in cancer target identification and drug discovery

“…Since network-based biology analysis applications for DTI prediction are usually based on the guilt-by-association principle 228 , it can only predict the interacting neighbors of known cancer targets. Currently, ML-based biology analysis applications can extend the predictions to downstream consequences 227 , thereby screening out more possible anticancer drugs.…”

“…Identifying drug-target interactions (DTIs) is crucial for drug screening. Especially, novel DTIs can be employed to look for the novel anticancer drugs with known targets 227 .…”

“…Currently, ML-based biology analysis applications have been employed to carry out drug screening and drug repurposing. For drug screening, previous studies have shown that network-based biology analysis applications can only screen the neighbour proteins of known targets, while drug-protein interactions may dysregulate the targets’ interacting neighbours 227 resulting in high false positive prediction results. ML-based biology analysis applications, such as graph-based neural network, have the advantage of integrated features that combine both ‘bottom-up’ 229 and ‘top-down’ 230 approaches to reduce the high false positive prediction results.…”

“…For example, Hinnerichs et al 227 developed the DTI-Voodoo that combines molecular features and phenotypes information with an interaction network using graph neural networks to predict drug-protein interactions (Fig. 11 ).…”

DTSyn: a dual-transformer-based neural network to predict synergistic drug combinations

“…Identifying chemical-gene interactions (drug-target interactions) is a crucial step in drug discovery or drug repurposing [Hinnerichs and Hoehndorf, 2021]. Finding new targets of approved drugs also help to analyze and identify new drug combinations as well as desirable therapeutic effects.…”