A computational walk to the hidden peaks of protein performance

Spiders use them to catch their prey, plants rely on them to fix carbon and mammals need them for eye vision—proteins.
Proteins play critical roles in nature, and not surprisingly, synthetic biologists heavily rely on their functional diversity to build new therapeutics (1), catalysts (2) and materials (3). But natural proteins are rarely optimal for their envisioned human uses. They rather need to be engineered to enhance their per- formance. Recently, researchers introduced a machine-learning guided paradigm that can predict which mutations in a pro- tein will enhance function with only 24 functional data sets as input (4). This paradigm could significantly accelerate the engi- neering of improved proteins for medicine, food, agriculture and industrial applications.