Since the field of protein structure prediction began in the 1960's, with the elucidation of the first protein structures, there has been much progress in understanding the forces and principles governing the folding of a polypeptide chain into a compact 3-D shape. However, judging by the latest results of the CASP structure prediction contest, this understanding is still insufficient to accurately predict the structure for a protein with no known structural homologues. We are working on an algorithm to predict contacts of unknown protein structures using only the sequence information present in the protein's family. Given that proteins evolve by natural selection, it is clear that random mutations in sequence are accepted only when they do not disturb the protein's functionality, or when they are compensated for by other mutations. A protein family, representing the same protein from different organisms, provides information about positions in the fold that have undergone concerted evolution. This information, present in coupled patterns of sequence change, reflects the residue's proximity in space.
Showing the most recent 10 out of 508 publications
