Computational protein design: software implementation, parameter optimization, and performance of a simple model

Computational protein design will continue to improve as new implementations and parameterizations are explored. An automated protein design procedure is implemented and applied to the full redesign of 16 globular proteins. We combine established but simple ingredients: a molecular mechanics description of the protein where nonpolar hydrogens are implicit, a simple solvent model, a folded state where the backbone is fixed, and a tripeptide model of the unfolded state. Sequences are selected to optimize the folding free energy, using a simple heuristic algorithm to explore sequence and conformational space. We show that a balanced parametrization, obtained here and in our previous work, makes this procedure effective, despite the simplicity of the ingredients. Calculations were done using our Proteins @ Home distributed computing platform, with the help of several thousand volunteers. We describe the software implementation, the optimization of selected terms in the energy function, and the performance of the method. We allowed all amino acids to mutate except glycines, prolines, and cysteines. For 15 of the 16 test proteins, the scores of the computed sequences were comparable to those of natural homologues. Using the low energy computed sequences in a BLAST search of the SWISSPROT database, we could retrieve natural sequences for all protein families considered, with no high-ranking false-positives. The good stability of the designed sequences was supported by molecular dynamics simulations of selected sequences, which gave structures close to the experimental native structure.