Improving Comparative Structure Modeling through the Use of Multiple Templates

A variety of methods are employed to model protein structures using multiple templates because they potentially provide more information than does the use of a single template. However, there has been little consensus on how best to use this added information, and not even a consensus that using multiple templates is always helpful. In this work we examine alternatives for integrating information from multiple templates and perform a systematic study to determine how using multiple templates affects model quality. Our data indicate that current methods appropriately weight the better template approximately 65% of the time, and that significant improvements are possible by using statistical potentials derived from a database of known structures to weight the contribution of each template to the final model structure. Funding: Burroughs‐Wellcome Predoctoral Fellowship and P01 GM071790