dc.description.abstract | Due to the excessive use of antibiotics, microbial pathogens have developed resistance to them, necessitating the urgent development of alternative therapies for infections. Antimicrobial peptides (AMPs) are small proteins that exhibit broad inhibitory effects against bacteria, fungi, parasites, and viruses et al. As a result, AMPs have emerged as a novel class of antimicrobial agents in recent years. In microbiology, the Minimum Inhibitory Concentration (MIC) refers to the lowest concentration that can inhibit bacterial growth and serves as an important indicator of drug activity. The primary objective of this study is to construct a regression model for predicting the MIC values of AMPs. Eight different model architectures were employed, along with various sequence features and genomic features, to assess the robustness of the frameworks. In this study, we ultimately utilized the contextual embeddings generated by a protein Language Model, combined with genomic features, in a deep learning architecture, achieving good evaluation results. Through an ensemble learning approach, the results of three top-performing supervised learning models were combined, and the ensemble model was evaluated. Pearson correlation coefficients of 0.756, 0.781, and 0.802 were obtained when testing the dataset against Staphylococcus aureus ATCC 25923, Escherichia coli ATCC 25922, and Pseudomonas aeruginosa ATCC 27853, respectively. These three strains are listed by the World Health Organization as requiring urgent research. These results demonstrate a certain level of accuracy in predicting the MIC using our ensemble model, which also exhibits good performance. | en_US |