Posters - Schedules

Antimicrobial peptides (AMPs), small polypeptide chains part of the innate immune response of multiple organisms, have emerged as a potential alternative to conventional antibiotics given their low propensity to generate bacterial resistance. Multiple artificial intelligence algorithms have been developed for the rapid discovery and design of powerful AMPs at low cost. However, these algorithms can be riddled with biases inherent in the databases used for their training. AI applications in the biological sciences often undermine such problems besides biological datasets being highly susceptible to bias for different reasons such as evolutionary redundancies or experimental conditions. Machine learning algorithms intertwine predictive and generative models to design optimized AMP sequences rationally. Despite a vast structural diversity, most considered AMP candidates fold into alpha-helical structures Simple observations or voluntary model bias, the limitations within these models lie in the diversity of peptide sequences and biological information. Here, we analyzed the structural landscape of GRAMPA, one of the largest repositories of AMPs sequences and related biological activities. We evaluated the performance of 13 state-of-the-art AMP predictive models against different structural classes looking for possible bias where we identified that structural class imbalance led to misclassified predictions.

Antimicrobial peptides (AMPs), small polypeptide chains part of the innate immune response of multiple organisms, have emerged as a potential alternative to conventional antibiotics given their low propensity to generate bacterial resistance. Multiple artificial intelligence algorithms have been developed for the rapid discovery and design of powerful AMPs at low cost. However, these algorithms can be riddled with biases inherent in the databases used for their training. AI applications in the biological sciences often undermine such problems besides biological datasets being highly susceptible to bias for different reasons such as evolutionary redundancies or experimental conditions. Machine learning algorithms intertwine predictive and generative models to design optimized AMP sequences rationally. Despite a vast structural diversity, most considered AMP candidates fold into alpha-helical structures Simple observations or voluntary model bias, the limitations within these models lie in the diversity of peptide sequences and biological information. Here, we analyzed the structural landscape of GRAMPA, one of the largest repositories of AMPs sequences and related biological activities. We evaluated the performance of 13 state-of-the-art AMP predictive models against different structural classes looking for possible bias where we identified that structural class imbalance led to misclassified predictions.