An integrated in silico screening framework for discovering novel antifungal peptide candidates using artificial intelligence and molecular docking.

Despite progress in antifungal therapeutics, invasive fungal infections are estimated to cause over one million deaths annually. In recent years, antimicrobial peptides (AMPs) have shown promise as a potential therapeutic option against such infections. Although over 2300 types of AMPs have been identified to date, the development of antifungal peptides (AFPs) has lagged behind. Therefore, the discovery of AFPs with strong efficacy against fungi remains a critical challenge. In this study, we developed an artificial intelligence system that learns from the amino acid sequences and physicochemical properties of known AFPs to efficiently identify putative AFP candidates. First, the amino acid sequences were effectively modeled using a combination of advanced algorithms from the field of natural language processing and a multi-layer perceptron. Second, the physicochemical attributes were learned using a model that combines Pfeature-derived features and a random forest classifier. By integrating these two models, we constructed a classification pipeline capable of rapidly identifying previously unreported putative AFP candidates from randomly generated amino acid sequences sampled according to empirical amino acid frequency distributions derived from known AFPs. Bioinformatic analyses suggested that these candidates exhibit structural and physicochemical features commonly associated with known AFPs and may potentially interact with enzymes involved in fungal cell wall biosynthesis. Future work will involve wet-lab validation of the antifungal activity, stability, and cytotoxicity of these putative AFP candidates to further assess the biological relevance and predictive utility of the proposed system.