Automatic Building of Protein Atomic Models from Cryo-EM Maps

Abstract

Cryo-electron microscopy (Cryo-EM) has emerged as a powerful method to obtain electron density maps of protein assemblies. However, reconstruction of atomic models from near atomic resolution (3-5 Å) cryo-EM maps is still challenging. Several tools developed for X-ray crystallography are widely used to interpret high resolution EM maps (around 3 Å), but poorly resolved side-chain densities hamper sequence attribution by automatic procedures at lower resolution. Furthermore, segmentation of EM maps into subunits remains a difficult problem when no structures of these subunits exist, or when conformational changes occur between the isolated and complexed form of the subunits. To tackle these issues, we have developed a graph-based method to thread most of the C-α trace of the protein backbone into the EM map. The EM density is described as a weighted graph such as the resulting minimum spanning tree encompasses the high density region of the map. A pruning algorithm is then applied to clean the tree and find the most probable positions of the C-α atoms, using sidechain density when available, in the form of C-α trace fragments. By complementing experimental EM maps with contact predictions from sequence co-evolutionary information, we demonstrate that our approach is able to correctly segment EM maps into individual subunits and recover most of the sequence attribution to generate full-atom models.