Ferguson and Wong publish PNAS paper on machine learning discovery and design of membrane active peptides

In a PNAS paper published online today “Mapping membrane activity in undiscovered peptide sequence space using machine learning” (http://dx.doi.org/10.1073/pnas.1609893113) Prof. Ferguson and Ben Fulan (Math, UIUC) report a support vector classifier as a new machine learning tool to efficiently screen peptide sequence space for membrane active peptides. Working in collaboration with Prof. Gerard Wong (BioE, UCLA) and his student Ernest Lee, experimental tests of peptides computationally designed using this approach validate that the classifier can discover new membrane active peptides, and also identify membrane activity in diverse and unexpected peptide families including neuropeptides, viral fusion proteins, topogenic peptides, and amyloids. This tool has implications for the rational design of peptides with diverse biomedical applications, including immunotherapy, intracellular drug delivery, broad-spectrum membrane-active antimicrobials, and nucleic acid transfection. The UIUC CoE press release is available here.