MUMBO: a protein-design approach to crystallographic model building and refinement

In recent years, significant progress has been achieved in automation of the crystal structure-determination process. However, the final part of this process, namely the refinement of the atomic model, is still tedious for biological macromolecules because, especially at lower resolution, it requires extensive manual intervention. Here, it is shown that computer algorithms widely used in protein-design approaches can substantially simplify this process, helping to identify the correct orientation of the side chains during refinement. This approach was implemented into the computer program MUMBO. As in many protein-design programs, side-chain rotamer diversity is generated using rotamer libraries. The selection of the best combination of side-chain orientations is based on either the dead-end elimination (DEE) theorem or a Metropolis Monte Carlo approach and on a detailed atomic scoring function that describes the molecular interactions between the rotamers. We show that this scoring function can be easily extended and complemented through the introduction of an X-ray pseudo-energy calculated from the electron density present at the position of the rotamer. This extension is fully compatible with present protein-design algorithms and it is shown for a number of test cases that using this approach model refinement is simplified and convergence occurs faster.