Comparative protein structure modeling as an optimization problem

Comparative or homology protein modeling uses experimentally determined protein structures to predict the conformation of another protein with a similar amino acid sequence. This technique can produce useful models for about an order of magnitude more protein sequences than there have been structures determined by experiment. Here, we review our approach to comparative protein modeling. In this approach, the three-dimensional model is calculated by satisfying spatial restraints extracted from an alignment of the sequence to be modeled with related known structures. We examine the types of errors in the resulting models and discuss some of the potential advantages of formulating comparative modeling as an optimization problem.