We present a molecular dynamics (MD) simulation study of the folding thermodynamics of thethree-stranded
-sheet protein Betanova. The protein and solvent are explicitly described by employing allatom models. An umbrella sampling technique was employed to probe thermodynamically relevant states atdifferent stages of folding. A database for the sampling was generated by conducting four high-temperaturesimulations. The initial conditions for the umbrella sampling were selected from this database of structures byemploying hierarchical clustering. Sampling of conformational space was then carried out at 275 K and thegenerated data were combined with the weighted histogram method to produce the two-dimensional foldingfree energy landscape. We found that the folding of the protein Betanova occurs in two collapse stages. Thefirst collapse brings the protein into a basin that contains various structures differing in their size and elementsof secondary structure. At the transition state from this basin of collapsed states to the native basin, the proteinadopts a native-like fold and size and forms
60% of native contacts. Thus the formation of native-like structureis concurrent with the secondary collapse. The overall stability of protein Betanova is found to be about 1kcal/mol, in agreement with the experimental estimate. We found that the native side chain contacts are theprimary factor in driving Betanova folding and stabilizing its native three-stranded
-sheet conformation. Bycontrast, hydrogen bonding is found to play a minor role in the folding of Betanova. Solvent is observed to bepresent in the protein core until late in folding.