Deriving Protein Structure Topology from the Helix Skeletion in Low Resolution Density Map using Rosetta

Electron cryo-microscopy (cryo-EM) is an experimental technique to determine the 3-dimensional structure for large protein complexes. Currently this technique is able to generate protein density maps at 6 to 9 Å resolution. Although secondary structures such as α-helix and β-sheet can be visualized from these maps, there is no mature approach to deduce their tertiary topology, the linear order of the secondary structures on the sequence. The problem is challenging because given N secondary structure elements, the number of possible orders is (2)*N!. We have developed a method to predict the topology of the secondary structures using ab initio structure prediction. The Rosetta structure prediction algorithm was used to make purely sequence based structure predictions for the protein. We produced 1000 of these ab initio models, and then screened the models produced by Rosetta for agreement with the helix skeleton derived from the density map. The method was benchmarked on 60 mainly alpha helical proteins, finding that for about 3/4 of all the proteins, the majority of the helices in the skeleton were correctly assigned by one of the top 10 suggested topologies from the method, while for about 1/3 of all the proteins the best topology assignment without errors was ranked the first. This approach also provides an estimate of the sequence alignment of the skeleton. For most of those true-positive assignments, the alignment was accurate to within +/2 amino acids in the sequence.