The accuracy of protein models automatically built into cryo-EM maps with ARP/wARP

Abstract

Recent developments in cryogenic electron microscopy (cryo-EM) have enabled structural studies of large macromolecular complexes at resolutions previously only attainable using macromolecular crystallography. Although a number of methods can already assist in de novo building of models into high-resolution cryo-EM maps, automated and reliable map interpretation remains a challenge. Presented here is a systematic study of the accuracy of models built into cryo-EM maps using ARP/wARP. It is demonstrated that the local resolution is a good indicator of map interpretability, and for the majority of the test cases ARP/wARP correctly builds 90% of main-chain fragments in regions where the local resolution is 4.0 Å or better. It is also demonstrated that the coordinate accuracy for models built into cryo-EM maps is comparable to that of X-ray crystallographic models at similar local cryo-EM and crystallographic resolutions. The model accuracy also correlates with the refined atomic displacement parameters.

Identification of unknown proteins in X-ray crystallography and cryo-EM

“…From the PDB we selected cryo-EM structures of ribosomes determined at a resolution better than 3.5 Å with half-maps available for download in the Electron Microscopy Data Bank (EMDB). Initially, all the structures were refined into corresponding maps with REFMAC5 using the auto_em.sh script from ARP/wARP 8.1 suite (Chojnowski et al, 2021). Out of these we selected refined models with CC_mask over 0.7 (estimated using phenix.map_model_cc, (Liebschner et al, 2019)).…”

“…As of 4 February 2020 we found 184 structures fulfilling these criteria. Initially, all the structures were refined into their corresponding maps with REFMAC5 (Murshudov et al, 2011a) using the auto_em.sh script from the ARP/wARP 8.1 suite (Chojnowski et al, 2021). Out of these, 117 models with CC_mask over 0.7 (estimated using phenix.map_model_cc, (Liebschner et al, 2019)) were selected for training the classifier.…”

“…To address these issues we developed findMySequence, a computer program that uses machine-learning predicted residue-type probabilities to query sequence databases using HMMER, a popular sequence analysis tool (Eddy, 2011). We show that the program successfully identifies sequences of protein models automatically built into crystallographic and cryo-EM maps, even though the automatically built models are usually highly fragmented and prone to tracing errors (Chojnowski et al, 2021). Furthermore, we use findMySequence to identify a protein purified and crystallized from a snake venom retrieved from Bothrops atrox, the most clinically important snake species in northern parts of South America (Estevao-Costa et al, 2016).…”

“…For cryo-EM, the sequence identification methods mentioned above in the context of X-ray crystallography are generally applicable. Nevertheless, cryo-EM maps are usually determined at resolutions lower than in crystallography, making the model building and the identification of side-chain identities far more ambiguous and challenging (Chojnowski et al, 2021).There are, however, examples of successful protein identification that avoid these issues by a bruteforce fitting of structures automatically predicted for a large pool of proteome sequences (Ramrath et al, 2018).…”

Cryo-EM single particle structure refinement and map calculation usingServalcat

“…This improvement in resolution has accelerated the development of methods for model building, refinement and validation. Automatic model building programs developed originally for crystallography are now being adapted for cryo-EM SPA maps (Terwilliger et al, 2018a;Hoh et al, 2020;Chojnowski et al, 2021). Density modification and local map sharpening can help interpret the map (Jakobi et al, 2017;Terwilliger et al, 2018b;Ramírez-Aportela et al, 2019;Ramlaul et al, 2019;Terwilliger et al, 2020).…”