Multiple protein folding nuclei and the transition state ensemble in two‐state proteins

Using exhaustive simulations of lattice models with side‐chains, we show that optimized two‐state folders reach the native state by a nucleation‐collapse mechanism with multiple folding nuclei (MFN). For both the full model and the Go version, there are certain contacts that on an average participate in the critical nuclei with higher probability than the others. The high‐ (≳0.5) probability contacts are largely determined by the structure of the native state. Comparison of the results for the full sequence and the Go model shows that non‐native interactions compromise the degree of cooperativity and stability of the native state. From an extremely detailed analysis of the folding kinetics, we find that non‐native interactions are present in the folding nuclei. The folding times decrease if the non‐native interactions in the folding nuclei are made neutral or repulsive. Using cluster analysis and making no prior assumption about reaction coordinate, we show that both full and Go models have three distinct transition states that give a structural description for the MFN. In the transition states, on an average, about two‐thirds of the sequence is structured, whereas the rest is disordered, reminiscent of the polarized transition state in the SH3 domain. Our studies show that Go models cannot describe the transition state characteristics of two‐state folders at the molecular level. As a byproduct of our investigations, we establish that our method of computing the transition state ensemble is numerically equivalent to the technique based on the stochastic separatrix, which also does not require a priori knowledge of the folding reaction coordinate. Proteins 2001;43:465–475. © 2001 Wiley‐Liss, Inc.