A probabilistic approach to protein backbone tracing in electron density maps

摘要: One particularly time-consuming step in protein crystallography is interpreting the electron density map; that is, fitting a complete molecular model of into 3D image produced by crystallographic process. In poor-quality maps, interpretation may require significant amount crystallographer's time. Our work investigates automating initial backbone trace maps. We describe ACMI (Automatic Crystallographic Map Interpreter), which uses probabilistic known as Markov field to represent protein. Residues are modeled nodes graph, while edges pairwise structural interactions. Modeling this manner allows be flexible, considering an almost infinite number possible conformations, rejecting any physically impossible. Using efficient algorithm for approximate inference—belief propagation—allows most probable protein's through map determined. test on set ten maps (at 2.5 4.0 Å resolution), and compare our results alternative approaches. At these resolutions, offers more accurate than current Contact: dimaio@cs.wisc.edu