Short-Range Contact Preferences and Long-Range Indifference: Is Protein Folding Stoichiometry Driven?

Hue Sun Chan
2011 Journal of Biomolecular Structure and Dynamics  
Mittal et al. (1) recently advanced an unconventional view on protein folding. By analyzing the spatial neighborhoods of amino acid residues in an extensive set of structures in the Protein Data Bank (PDB), the authors concluded that preferential interactions between amino acid residues do not drive protein folding. In this connection, it should be noted that preferential interactions between amino acids are the basis for introducing knowledge-based potentials, which in turn provide the
more » ... ning for present day three-dimensional protein structure prediction by modeling and simulation (2-5 and references therein). Instead of these preferential interactions, Mittal et al. indicate that "protein folding is a direct consequence of a narrow band of stoichiometric occurrences of amino-acids in the primary sequences" (1). According to the authors, this observation is akin to Chargaff's discovery that the molar ratios of adenine and thymine and that of guanine and cytosine in DNA were not far from unity (6) . This assertion runs counter to prevalent views, most notably the decades-old consensus that hydrophobic interactions is a major driving force for folding (7, 8) . The view of Mittal et al. is counterintuitive because folded proteins do have a "hydrophobic inside, polar outside" organization; the average buried area (not exposed to solvent) of an amino acid residue in folded proteins correlates with its hydrophobicity (9). The authors' conclusion is all the more puzzling in light of established statistical potentials derived from the PDB that clearly demonstrate preferences in contacts among amino acids (10-12). A major contribution to those preferences is none other than the hydrophobic effect (13). The conclusion of Mittal et al. was based on enumerating the spatial distribution of pairs of Cα positions among PDB structures. For each of the 20 × 20 pairs of the twenty types of amino acids, they obtained the number of residue pairs (termed "contacts") within a variable distance from each other (the residues were referred to as "neighbors" regardless of distance), and fitted the distance dependence of the number of such contacts to a particular sigmoidal-shaped function. They found that the fitted sigmoidal trends were similar for all 20 × 20 types of neighbors, and that asymptotically (at large distances) the number of contacts of an amino acid type is proportional to its overall composition in the PDB structures considered. They interpreted the results of this "neighborhood analysis" of theirs (1) as implying a lack of preferential interactions. Mittal et al. did not address the inconsistency of their conclusion with established statistical potentials. But this contradiction is significant because it should not have arisen. After all, the authors' results and the statistical potentials were both derived from the PDB.
doi:10.1080/07391102.2011.10524960 fatcat:hvbnx5yvqjcqnox43pzeyme5aq