Associative learning mechanism for drug‐target interaction prediction

As a necessary process of modern drug development, finding compound that can selectively bind to specific protein is highly challenging and costly. Exploring drug-target interaction strength in terms affinity (DTA) an emerging effective research approach for development. However, it model interactions deep learning manner, few studies provide interpretable analysis models. This paper proposes DTA prediction method (mutual transformer-drug target [MT-DTA]) with interactive autoencoder mechanism. The proposed MT-DTA builds variational autoencoders system cascade structure the attention convolutional neural networks. It not only enhances ability capture characteristic information single molecular sequence but also establishes expression relationship each substructure sequence. On this basis, module constructed, which adds paths between pairs complements correlations substructures. performance was verified on two public benchmark datasets, KIBA Davis, results confirm predicting DTA. Additionally, transformer models different configurations improve feature drug/protein molecules. performs better correctly strengths compared state-of-the-art baselines. In addition, diversity molecules be expressed than existing methods such as SeqGAN Co-VAE generate more new drugs. value fuses pair output predicted theoretically proves maximises evidence lower bound joint distribution model, consistency probability actual values. source code available at https://github.com/Lamouryz/Code/tree/main/MT-DTA.