Introduction

D3AI-CoV is a deep learning platform for predicting drug targets and for virtual screening against COVID-19. It was developed with two purposes, one is for predicting target proteins for active compounds observed from experimental studies, and another is for virtual screening based on deep learning-based models. To this end, we expanded database of the active compounds against 9 pathogenic coronaviruses. Based on the database, we utilized two approaches to construct two different models for predicting the targets, namely MultiDTI and MPNNs-CNN and one models, namely MPNNs-CNN-R, for virtual screening against COVID-19. “TargetPrediction” is for predicting target proteins, while “VirtualScreening” is for virtual screening against target proteins based on the MPNNs-CNN-R model.

MultiDTI model. A three-layer convolutional neural network and multiple down-sampling residual layers are used to extract features of SMILES and protein amino acid sequence, and multilayer perceptron is used to project the representations of drugs and targets into the common space.

MPNNs-CNN and MPNNs-CNN-R models. MPNNs and CNN are used to extract features of compound SMILES and protein sequence, respectively. Multilayer perceptron and logistic algorithm were used to predict potential connections between compounds and targets.