Protein design by citizen scientists

Abstract

Proteins are a class of molecule best known for their tendency to fold into well-defined 3-dimensional structures. The structure of a protein is determined by the sequence of amino acid units that make up the protein. Our understanding of the sequence-structure relationship has recently reached the point that we can design new proteins de novo (i.e. without reference to existing protein sequences). However, this understanding is only partially encoded in protein design software, which still requires a user with considerable expertise in protein engineering. Here, I use citizen science to identify and resolve limitations of protein design software, by crowdsourcing protein design tasks to non-experts playing the computer game Foldit. Using the output of Foldit players as feedback, I iteratively trialed and improved protein design software to the point that non-experts can now use the software to successfully design proteins from scratch. This work reveals implicit assumptions of expert protein engineers, corrects errors in the Rosetta protein structure energy function, and shows how citizen science can be used to improve a scientific model.