Modeling of Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) Proteins by Machine Learning and Physics-Based Refinement

Lim Heo; Michael Feig

We evaluated MolProbity score 11 for all available models ( Table 3) . Although this score does not indicate whether a model reflects the overall correct structure or not, it can tell if a given model satisfies basic protein stereochemistry or not. All of the models generated from our modeling pipeline has less than 1.5 MolProbity score. Especially, steric clashes and rotamer outliers, rarely exist. Most of the AlphaFold models also have good MolProbity scores, although there are sometimes a few numbers of steric clashes between atoms and some rotamer outliers. After refining those models, most of the poor local geometries could be resolved, and the resulting MolProbity scores after refinement of the AlphaFold models are very good. In contrast to these models, models from Zhang group have poor local geometries as measured by the MolProbity score. These models have numerous atomic clashes, poor side-chain conformations, and bad backbone dihedral angles that generally suggest poor stereochemistry. The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2020.03.25.008904 doi: bioRxiv preprint The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2020.03.25.008904 doi: bioRxiv preprint The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2020.03.25.008904 doi: bioRxiv preprint