High-Throughput Protein Design Reveals Quantitative Protein Stability Requirements

Abstract

Despite two decades of progress in computational protein design, a large fraction of de novo designed proteins fail to fold as designed due to our incomplete understanding of the principles governing protein stability. These challenges persist in part because design studies to date have been limited to testing at most tens of proteins, which provides insufficient experimental data to identify causes of failure or to improve quantitative modeling. To overcome this, we combined computational protein design with next-generation oligo library synthesis and a high-throughput protease susceptibility assay to measure stability for thousands of unique designed small proteins and control sequences simultaneously. We identified hundreds of new stable designs across four targeted topologies, and a subset of these were tested individually and found to be monomeric, fold cooperatively with high stability for their small size, and to form structures in solution closely matching their designed structures. Iterating between high-throughput design and testing enabled us to statistically examine hypotheses about protein stability and led to improved design success rates, including for topologies with no successes in the first attempt. The stabilities of the more than 10,000 designed proteins examined here provide the most detailed picture yet obtained of the stability requirements for small proteins, and this approach promises to change computational protein design from low-throughput craft into data-driven engineering.