Deep learning-based local quality estimation for protein structure models from cryo-EM maps

Abstract

In the recent decade, an increasing number of cryo-electron microscopy (cryo-EM) maps were determined to model the tertiary structures of biomolecules. Although the resolution of determined cryo-EM maps is improving in general, there are still many cases where amino acids of a protein were assigned to the map with different confidence levels, including those with relatively high ambiguity. A common error, particularly when side-chain density is not well defined and there are poorly ordered loops between secondary structures, is for the sequence of the model to become out of register.