MDAD/ABiofilm

Numerous studies have demonstrated that microbes play a vital role in human health, making the identification of potential microbe-drug associations critical for drug discovery and clinical treatment. In this manuscript, we proposed a novel prediction model named GTDEKAN by integrating an aware Transformer network with a Dual Cross-Attention (DCA) module (including a Channel Cross-Attention and a Spatial Cross-Attention) and an Enhanced Kolmogorov-Arnold Network (EKAN) to infer potential microbe-drug associations. In GTDEKAN, we first constructed a heterogeneous microbe-drug network N through combining multiple similarity metrics of microbes, drugs, and diseases. Next, the aware Transformer framework, combined with the DCA module, would be applied to extract global and local features of nodes in N. Finally, we would further adopt the EKAN obtain the predicted scores of potential microbe-drug associations. Additionally, in order to evaluate the predictive performance of GTDEKAN, intensive experiments have been conducted, and results show that GTDEKAN can achieve reliable AUC and AUPR values of 0.9859 and 0.9733 respectively, which outperform state-of-the-art competitive methods. And simultaneously, case studies of four well-known drugs across different databases demonstrate the effectiveness of GTDEKAN as well, which underscores its potential to uncover unknown microbe-drug associations.