Project I Improved algorithms for structure refinement at multiple resolution ranges Our ability to visualize macromolecules at atomic resolution has had a lasting and profound impact on the study of biology. Obtaining the most accurate model is paramount for maximizing biological insight and for enabling activities such as the development of new therapeutics for human health. For the most challenging, and often very biologically important systems the availability of only low resolution diffraction data (3-4.5 Å) has limited the accuracy of models. Therefore we will develop computational methods that significantly improve models, by creating better methods for their refinement even when data are limiting. This will have a broad impact on structural biologists, drug developers and molecular modelers. The field of X-ray crystallography has matured significantly over the last 10 years. However, there still remain many challenges in solving and refining structures at low resolution. These same challenges are now being encountered by researchers in the field of single particle cryo-EM. Many of the tools we have developed for crystallography can be applied to this problem. Therefore we propose to further develop our methods to address structure refinement when only sparse experimental data is available. Real space algorithms will be implemented to improve the refinement of models against maps, in particular for the cryo-EM case. The resulting procedure will be able to robustly refine the fit of models to maps while preserving excellent stereochemistry. Obtaining accurate stereochemistry has profound implications for the understanding of enzymatic reactions, macromolecular interactions, and the development of novel therapeutics. We will develop methods to generate the highest quality molecular models, which provide the greatest chemical and biological insight, at any resolution range. To do this we will continue our collaborations that have made use of methods from the field of structure modeling, and extend them to make use of other physically reasonable molecular potentials developed in the fields of molecular mechanics and dynamics. Finally, we will continue to develop and maintain the Phenix infrastructure to support both novice and expert researchers. Diffraction image analysis will be incorporated into the Phenix system to enable automated analysis of user data to detect possible problems. We will also implement a structure deposition system that incorporates model and data validation, and brings together information from multiple stages of structure solution.