Predicting activatory and inhibitory drug–target interactions based on mol2vec and genetically perturbed transcriptomes

Lee, Won-Yung; Lee, Choong-Yeol; Kim, Chang-Eop

A computational approach to identifying drug–target interactions (DTIs) is a credible strategy for accelerating drug development and understanding the mechanisms of action of small molecules. However, current methods are limited to providing simple DTIs without mode of action (MoA), or they show unsatisfactory performance for a limited range of compounds and targets. Here, we propose AI-DTI, a novel method that predicts activatory and inhibitory DTIs by combining the mol2vec and genetically perturbed transcriptomes. We trained the model on large-scale DTIs with MoA and found that our model outperformed a previous model that predicted activatory and inhibitory DTIs. To extend the applicability, we applied the inferential method for the target vector so that our method can learn and predict a wider range of targets. Our method achieved satisfactory performance in an independent dataset where the target was unseen in the training set and a high-throughput screening dataset where positive and negative samples were explicitly defined. Finally, our method successfully rediscovered approximately half of the DTIs for drugs used in the treatment of COVID-19. These results indicate that AI-DTI is a practically useful tool for guiding drug discovery processes and generating plausible hypotheses that can reveal unknown mechanisms of drug action.