Improved methods for the annotation and processing of Cryo-EM images and for atomic model validation

Cryogenic electron microscopy was introduced in the 1980s as an innovative technique to determine protein structures and has become more popular in recent years. With almost 7800 maps released last year, it is the second most used method after X-ray crystallography. The technological advancements and the constant improvements in automated data collection create a strong need for software tools to automatically assess the quality of the collected data.
In this work, we identified the parts of the cryo-EM processing pipeline, from data collection to atomic model building and validation, that can significantly benefit from automated tools. The evaluation of automated particle picking tools revealed a common issue when the picking is less effective from micrographs with non-uniform ice distribution. We developed a software tool which could mitigate this problem and equalise the contrast locally. Additionally, we introduce a new parameter for the processing, which can be used to associate the particles' coordinates with the estimated ice thickness levels, which can also be calibrated to the measured thickness. This software is added to the ISpyB data collection pipeline at the Electron Bio-Imaging Centre in Diamond Light Source.
A software tool was developed for atomic model validation based on the False Discovery Rate approach which allows scoring each residue in the model by checking if they are placed within the cryo-EM map density or in the background noise. This tool is now available from the Collaborative Computer Project for Cryo-EM (CCPEM) software suite.