Peptide design by optimization on a data-parameterized protein interaction landscape

Significance Medicine, agriculture, and the biofuel industry use engineered proteins to perform functions such as binding, catalysis, and signaling. Designing useful proteins faces the “needle in a haystack” problem posed by the astronomical number of possible sequences. Proteins of utility can be found by experimentally screening 102 –109 molecules for properties of interest. We posit that such screens can serve as the beginning of a powerful computationally aided design process. Data collected in high-throughput experiments can be used to learn aspects of the relationship between protein sequence and function. We show how models trained on data can guide computational exploration of huge sequence spaces. This can enable rational design of molecules with custom properties that would be difficult to discover using other techniques.