Comment on the Communication TMThe Key to Solving the Protein-Folding Problem Lies in an Accurate Description of the Denatured State∫ by van Gunsteren et al.**

In a recent communication, [1] van Gunsteren et al. proposed that the folding of proteins is much easier than has been believed [2] because the number of accessible conformers in the denatured state is only on the order of 10 9 for a 100-residue protein, instead of on the order of 10 24 to 10 30 used in previous analyses. [3, 4] This would make it possible for a polypeptide chain to fold to the native state in about 1 s by a random search since the transition from one conformer to another requires about 1 ns. [5] Van Gunsteren et al. [1] based their argument on simulations ranging from 32 to 200 ns of the folding and unfolding of six b and a amino-acid peptides in methanol and water. For these peptides, the number of conformers found by them corresponds to 1.22 to 1.74 con-formers per torsional degree of freedom, in the same range as the value 1.33 we found [6] in Monte Carlo simulations with an implicit solvent model of a denatured 16-residue peptide that forms a b-hairpin in the folded state; the simple average of the number of conformers per torsion angle calculated by van Gunsteren et al. for the different peptides is 1.40. If the number of conformers scales exponentially with the number of amino acids, an estimate of 1.33 (1.40) conformers per backbone torsion angle leads to a denatured state for a 100-residue protein with 1.33 200 (1.40 200) ˆ 6  10 24 (1.7  10 29) accessible conformers, [3, 4] which is many orders of magnitude larger than the value (10 9) suggested by van Gunsteren et al. The essential question concerns not the number of conformers observed in the peptide studies, but the statement that the number of conformers ™does not grow exponentially with the length of the peptide∫ [p. 364/352 of ref. [1]]. It is the latter that leads to their optimistic conclusion concerning the simplicity of the protein folding problem. There is no evidence for or against this claim in their simulations. The peptides studied vary by only a factor of two in size; in fact, there is one short peptide (9 torsion angles) and the five others vary in length only between 16 and 21 torsion angles. Since the number of conformers shows essentially no correlation with the number of torsion angles (the Pearson linear correlation coefficient between the …