The function of a protein is generally determined by its three-dimensional (3D) structure. Thus, it would be useful to know the 3D structures of the thousands of protein sequences that are emerging from the many genome projects. This is the aim of structural genomics. The aim will be achieved by a focused, large-scale determination of protein structures by X-ray crystallography and nuclear magnetic resonance spectroscopy, combined efficiently with accurate protein structure modeling techniques. In particular, comparative or homology-based protein structure modeling is expected to play a major role in this effort. Comparative modeling calculates a 3D model of a given protein sequence from the previously determined structures of related proteins. It involves fold assignment, sequence-structure alignment, model building, and model evaluation. To enable large-scale modeling, these steps are being assembled into a completely automated pipeline. The methods involved in the pipeline and their performance are reviewed. The errors in the resulting models are described and their uses in biology are discussed.