The inverse protein folding problem: Protein design and structure prediction in the genomic era

Millions of proteins are being identified every year by high throughput genome sequencing projects. Many others can potentially be created by protein engineering and design methods. Here, we review a method for computational protein design (CPD), which starts from a known protein and its 3D structure, and seeks to modify it by mutating some or all of the amino acid sidechains. The mutations are selected to provide stability, and possibly other properties, such as ligand binding. For each set of candidate mutations, the 3D structure is modeled, with an assumption of small, localized perturbations; in particular, we assume the backbone conformation does not change significantly. As in other CPD implementations, the structure is modeled using a classical, molecular mechanics approach along with a simple, implicit description of solvent. Some of the calculations have been distributed to volunteers on the Internet, through our Proteins@Home volunteer computing project. The method and selected results are described, which show that the designed sequences share important properties of natural proteins.