The long coming of computational structural biology

Fifty years ago, the structures of the alpha-helix, the beta-sheet, the alpha-helical coiled coil and the collagen triple helix raised the expectation that protein structure could be understood computationally, using a combination of geometric considerations, model-building and parametric equations. The first crystal structures dispelled this hope, revealing a disconcerting lack of regularity in the folding patterns of proteins. Gradually it became clear that the protein folding problem-namely deducing the structure of a protein from its amino acid sequence-was of exceptional difficulty. Its solution has remained outside our reach to this day and, arguably, it represents the most important unsolved problem in molecular biology. Nevertheless, our ability to understand and predict molecular structure by computational means has made steady progress, suggesting that we will eventually conquer the problem, not by a few heroic insights, but by steady advances in biophysical knowledge, biological databases, software applications and raw computer power. Computational structural biology, whose influence is already pervasive, will come to dominate structural approaches in the next decades.