Modern triple resonance protein NMR backbone assignment, using AlphaFold and unlabelling to drive chemical shifts assignment in proteins

NMR is a powerful technique to study the structure, dynamics and interactions of proteins. However, to obtain atomic resolution data, NMR signals must first be correlated with specific chemical groups – a problem called assignment. The software AlphaFold has shown to be a great advancement in modern-day science. Until now, structural analysis of proteins had been bottlenecked by months/ years’ worth of slower techniques that were traditionally used to determine a protein structure In this project we have designed and applied a semi-automated assignment program called SNAPS (Simple NMR Assignment using Predicted Shifts). This allows the user to go from a set of NMR spectra of a protein with a known 3D AlphaFold or X-ray crystallography structure to a fully assigned chemical shifts of the backbone resonances. In addition, unlabelling experimental data can be incorporated into the use of the program to generate more reliable assignment data by helping the program along in the mapping of the amino acids for assignment by providing places in the assignment where the amino acid it is known. The program was largely written by Dr Alex Heyam from the university of Leeds but testing scripts and well as a NEF importer was written to ensure the program was user-friendly and ensured rigid, fool-proof datasets being imported for backbone assignment. The program also underwent large-scale testing on roughly 150 proteins with known 3D crystal structures as well as 15N unlabelleing data being implemented to improve assignment. AlphaFold structures could also be implemented for use in the program also. Assignment was as good as 86-88% dependent upon parameters, with the unlabelling being slightly more effective with assignment (for PDB crystal structures).