Deep learning integration of molecular and interactome data for protein-compound interaction prediction

Motivation: Virtual screening, which can computationally predict the presence or absence of protein-compound interactions, has attracted attention as a large-scale, low-cost, and short-term search method for seed compounds. Existing machine learning methods for predicting protein-compound interactions are largely divided into those based on molecular structure data and those based on network data. The former utilize information on proteins and compounds, such as amino acid sequences and chemical structures, and the latter utilize interaction network data, such as data on protein-protein interactions and compound-compound interactions. However, few attempts have been made to combine both types of data in molecular information and interaction networks. Results: We developed a deep learning-based method that integrates protein features, compound features, and heterogeneous interactome data to predict protein-compound interactions. The interactome data consist of protein-protein interactions and compound-compound interactions. The performance evaluations show that our deep learning framework for integrating molecular structure data and interactome data outperforms state-of-the-art machine learning methods for protein-compound interaction prediction tasks. This reveals that protein-protein interaction and compound-compound interaction networks capture different perspectives than amino acid sequence homology and chemical structure similarity, and they have a synergistic effect in improving prediction accuracy. Furthermore, experiments on three datasets of different difficulties show that this method is more robust than existing methods in accurately predicting interactions between proteins and compounds that are unseen in the training samples.