The VUStruct Pipeline is moving to Leipzig University. Unfortunately, web input of variants will not be possible during this transition.
The VUStruct backend will continue to run and we'd still welcome the opportunity to process variants for you. Please reach out to discuss your project.
Thanks for all the collaborators who made this possible. Please now cite: VUStruct: A compute pipeline for high throughput and personalized structural biology
If you have any difficulties using VUStruct, please Get in touch.Many proteins function through binding to other proteins partners, and approximately 60% of disease-associated missense mutations have been noted to perturb PPIs(Sahni et al. 2015). Amino acid variants in interaction surfaces could have negligible impact in folding energetics and still be deleterious through disruption of protein binding to usual partners. The ScanNet(Tubiana et al. 2022) machine learning algorithm was trained through analysis of the Dockground(Kundrotas et al. 2020) database of 3D protein-protein interacting structures. Operationally, for each variant position covered by an Alphafold model, we ask ScanNet to predict whether the position participates in a PPI binding surface.
We report any variant position to which ScanNet assigns at least 50% probability of being involved in a PPI.