Model and density from data set 6 using an energy-optimized model as the source of phase information. Reprinted from Nature. 1
Molecular replacement is a method for solving the phase problem of X-ray diffraction data sets. The approach relies on the existence of a known protein structure with some homology to the unknown structure, but it is not very successful with low levels of homology (less than 30% sequence identity). However, even poor initial electron density maps are rich with information, and Baker and colleagues (PSI JCSG and NESG) have now applied Rosetta, an energy-guided optimization method, to hone initial models generated from these maps. These improved molecular-replacement models were then passed into crystallographic model building techniques and Rfree used to assess the success of the method. The data were also assessed by the groups that provided the starting intractable data sets. The authors examined 18 data sets that were refractory to solution by previous methods and, of the 13 that could not also be solved by other methods, 8 were ultimately considered solved using the Rosetta refinement method. The authors extensively compared the method to others, including simulated annealing, arguing that it consistently performed strongly. In addition, by examining performance with comparison to homologs of differing sequence identity, the method was found to perform well above 20% identity. Although the method requires that there be fewer than four copies in the asymmetric unit and that the resolution be 3.2 Å or better, it promises to greatly increase the set of structures that are in the realm of solution by molecular replacement.