(449f) Classifying Antimicrobial and Multifunctional Peptides with Machine Learning

Barrett, R. - Presenter, University of Rochester
White, A., University of Rochester
Antimicrobial and antifouling peptides are highly desirable biomolecules, but they are relatively rare among all known peptides. It is not intuitively apparent from an amino acid sequence whether a given peptide will exhibit these properties. Thus, it is of interest to develop a data-driven quantitative structure-activity (QSAR) model to predict potentially antimicrobial and antifouling peptides. Machine learning techniques are well-suited to such a task, and have the advantage of a robust, established pool of algorithms for learning. Here, we exhibit Bayesian network QSAR models that allow us to accurately predict whether an input peptide may exhibit antimicrobial properties, given its sequence.