Assembly of Protein Tertiary Structures

In the last several years there has been exciting progress in the development of algorithms for ab initio protein folding: the generation of protein structures starting from amino acid sequence infor- mation alone (Kolinski & Skolnick, 1994; Bowie & Eisenberg, 1994; Yue & Dill, 1996, Srinivasan & Rose, 1995). Because of the many formidable pro- blems facing ab initio folding simulations, such al- gorithms are not likely to become useful methods of structure prediction for any but the smallest pro- teins for quite some time. However, such efforts are of crucial importance because they highlight, as few other experiments can, the challenges facing current studies of protein folding. Our primary interests in this area are twofold: ®rst, to develop a computational model to com- plement biophysical and molecular biological stu- dies of the folding of very small protein domains currently underway in our laboratory, and second, to build upon our studies of local sequence-struc- ture relationships (Han & Baker, 1996), which are currently limited by a lack of treatment of non- local interactions. A working model for protein folding is that local amino acid sequence propensi- ties bias each local segment of a folding polypep- tide chain towards a small number of alternative local structures and that non-local interactions preferentially stabilize native-like arrangements of these otherwise transient local structures. The weak nature of the local propensities has compli- cated both the prediction of secondary structure from sequence and the search for structure in pro- tein fragments (Bystroff et al., 1996). Here, we use a knowledge-based treatment of local interactions related to that of our previous studies: short se- quence segments are restricted to the local struc- tures adopted by the most closely related sequences in the protein structure database.

Free download research paper