Automated refinement for protein crystallography: LAFIRE

Structural refinement of macromolecules is difficult because the observation-to-parameter ratio (typically about 3-5) is usually low. In most cases, manual intervention is required as a complementary function to the refinement algorithms; linking and/or extending the fragments of the initial model and fitting ill-matched residues using computer graphics software. Those manual steps consume time and require a great deal of expertise in crystallography. From the viewpoint of image processing and pattern recognition, the linking and/or extending process can be considered as a problem of looking for a path between two known nodes in a graph, while the fitting process can be considered as an exhaustive search of the parameter space in a local region. Moreover, combining the solution of these two problems with an expert refinement protocol, the automation of refinement can be achieved (LAFIRE).

In my talk, I will present the detailed algorithms designed for solving these problems and the expert refinement protocol employed by LAFIRE. Also, some examples (with their specific aspects) refined automatically will be shown. In addition, I will be available to guide the students how to refine a sample or their own structure with specific problems automatically through LAFIRE.