| Folding, the spontaneous disorder to order transition of protein molecules under physiological conditions, is the result of interplay between amino acid sequence and physical forces that drive the protein's free energy to a minimum. Firm understanding of the nature of this interplay would be demonstrated by consistent successful first-principles prediction of a wide variety of protein structures from their sequences alone. This challenge, known as the protein folding problem, remains one of the most fundamental unanswered questions in biology. However, much progress has been made in protein structure prediction through an empirical method known as fragment assembly, in which unrelated segments of previously-solved protein structures are randomly stitched together to recover the native structure of the protein in question. These segments exhibit considerable conformational degeneracy, suggesting local conformational bias, but no physical explanation for this bias has yet been offered.;The object of my thesis is to characterize prominent local motifs in protein fragments and then determine the physical-chemical factors responsible for them. As regular secondary structure (alpha-helices and beta-strands) has been studied extensively, I focus my attention on coil segments of proteins in three studies. My findings are as follows. First, over 90% of protein structure is covered by only 7 local motifs, all of which can be reproduced in simple Monte Carlo Simulations imposing only steric restrictions and hydrogen bond satisfaction, which requires all polar groups in a protein have a hydrogen bond partner from solvent or another intrapeptide polar group. Second, hydrogen bond satisfaction systematically winnows available conformational space in the unfolded state, limiting conformational variability. Finally, 98% of prominent short (2- to 4-residue) turn conformations can be reproduced in simple simulations imposing only steric restriction and hydrogen bond satisfaction and rewarding intrapeptide hydrogen bonds. The library of turn conformations resulting from these simulations is sufficient to cover over 90% of 5-20 residue coils. Together these three studies demonstrate the critical role of sterics, hydrogen bond satisfaction and hydrogen bonding in selecting local biases in protein structure and suggest next steps for generating a protein fragment library from physical-chemical first principles. |
