PDF《A biophysical protein folding》

s overall fold, and mutations at these sites mainly alter stability but not activity. The stability changes induced by those mutations are predictable in a statistical sense, as explained in Results. Though less predictable than stability, activity is governed mostly by just a few key residues, e.g. the active catalytic site, where mutations nearly always abolish activity. This general picture of protein organization underlies our biophysics-based model for approximating fitness effects of mutations. Though limited in its ability to describe all mutations, here we argue that our model provides the first simple, bottom-up approach to understanding mutational fitness effects. We strengthened our protein model by merging it with stochastic population genetics simulations that include polyclonality, genetic drift, and linkage between sites. In our simulations, model proteins are continually buffeted by mutations that usually undermine protein stability. These destabilizing mutations shift the folded-unfolded equilibrium toward unfolded proteins, which imposes a context-dependent, usually small, fitness penalty proportional to the extent of unfolding. Partially destabilized proteins can be compensated by subsequent stabilizing mutations that replenish the fraction of folded protein and improve fitness. The asexual population dynamics yields a steady state distribution of protein stabilities (p(?G)), from which we obtain a DFE with the same qualitative shape as observed experimentally. We find that p(?G) shifts toward instability for high mutation rates and small population sizes. This shift in stability disperses the DFE, i.e. increases the absolute selection coefficient of mutations and decreases robustness. Although the principles of our model are applicable to all species, we focus on viruses due to their relative simplicity and their extensively measured DFE (7-11). Results Nearly neutral thermodynamic fitness landscape. Model genomes comprise ? genes, each encod........