Artificially designed small peptides are currently quite interesting for medical research since they provide a way to target specific cellular activities. In many cases one wants to combine several functions into a single peptide that is as small as possible. This is problematic as the activity of the peptide is often lost when combining several functions.
In our article we propose a design strategy based on “compatible function”, meaning functions that require similar physio-chemical properties of the peptide. Starting from a peptide template already containing two compatible functions (DNA-binding and pheromone activity) we use a combination of Machine learning and optimization to add two additional functions (cell penetration and antibacterial) for the entire peptide by adding only a few amino acids. The resulting peptides are all very active and provide antibacterial activity on the level of Ampicillin with one peptide being even being much more efficient than Ampicillin.