Multi-Source Language Models Fusion with Dynamic Routing Capsule Network for Antimicrobial Compound Identification

Antibiotics are a vital class of drugs closely linked to the prevention and treatment of bacterial infections. Accurate prediction of molecular antimicrobial activity remains a key challenge in the discovery of novel antibiotic candidates. Laboratory-based identification of antimicrobial compounds is costly, time-consuming, and often results in rediscovery of known antibiotics.

In this study, we propose a model that incorporates a capsule network architecture and introduces innovations in loss-function selection and feature-processing modules. The model demonstrates superior performance in predicting inhibitory activity against Escherichia coli, Acinetobacter baumannii, and Staphylococcus aureus, and outputs the probability that a given compound exhibits antibacterial activity against these two strains.

To improve accessibility, we have developed an intuitive web portal for the model, available at https://dmci.xmu.edu.cn/CapMolPred/indexpage.php.