Automation and assessment of de novo modeling with Pathwalking in near atomic resolution cryoEM density maps

With the rapidly growing number of macromolecular structures solved to near-atomic resolution using electron cryomicroscopy (cryoEM), map interpretation and model building directly from the density without the use of structural templates has become increasingly important. As part of the 2015/2016 Map and Model Challenge, we attempted to assess our latest de novo modeling tool, Pathwalking, in terms of performance and usability, as well as identify areas for future improvements. In total, we applied Pathwalking to six density maps between 3 and 4.5 Å resolution selected from the challenge data sets. In five of the six cases, Pathwalking was able to accurately determine the protein fold and in three of these cases, the final all atom model had less than 1.6 Å RMSD when compared to the known structure. Model building and refinement was nearly completely automated, used default parameters and took less than 30 min to complete a refined all atom model. A direct outgrowth of this work was a more streamlined automated command line Pathwalking utility, as well as a novel sequence assignment and optimization routine, which attempts to register sidechain density with expected side chain volume. In total, Pathwalking offers a nearly complete, robust and efficient method for constructing atomistic protein structures directly from a density map without the aid of a template.