Post-translational modifications to proteins are often critical to function. However, these modifications (phosphorylation, glycosylation, and others) are typically unresolved in experimental X-ray structures, and never seen in model structures. Sequence-based prediction of post-translational modification sites has been evolving in recent decades, and we have integrated MutsiteDeep (Wang et al. 2017), the first deep-learning framework for prediction of these sites.
We report variants at sites predicted to have at least a 50% likelihood of post-translational modification.