AttenSyn: An Attention-Based Deep Graph Neural Network for Anticancer Synergistic Drug Combination Prediction

Identifying synergistic drug combinations is fundamentallyimportantto treat a variety of complex diseases while avoiding severe adversedrug-drug interactions. Although several computational methodshave been proposed, they highly rely on handcrafted feature engineeringand cannot learn better interactive information between drug pairs,easily resulting in relatively low performance. Recently, deep-learningmethods, especially graph neural networks, have been widely developedin this area and demonstrated their ability to address complex biologicalproblems. In this study, we proposed AttenSyn, an attention-baseddeep graph neural network for accurately predicting synergistic drugcombinations. In particular, we adopted a graph neural network module to extract high-latent features based on the molecular graphs onlyand exploited the attention-based pooling module to learn interactiveinformation between drug pairs to strengthen the representations ofdrug pairs. Comparative results on the benchmark datasets demonstratedthat our AttenSyn performs better than the state-of-the-art methodsin the prediction of anticancer synergistic drug combinations. Additionally,to provide good interpretability of our model, we explored and visualizedsome crucial substructures in drugs through attention mechanisms.Furthermore, we also verified the effectiveness of our proposed AttenSynon two cell lines by visualizing the features of drug combinationslearnt from our model, exhibiting satisfactory generalization ability.