UCBShift 2.0: Bridging\nthe Gap from Backbone to Side\nChain Protein Chemical Shift Prediction for Protein Structures

Chemical shifts are a readily obtainable NMR observable\nthat can\nbe measured with high accuracy, and because they are sensitive to\nconformational averages and the local molecular environment, they\nyield detailed information about protein structure in solution. To\npredict chemical shifts of protein structures, we introduced the UCBShift\nmethod that uniquely fuses a transfer prediction module, which employs\nsequence and structure alignments to select reference chemical shifts\nfrom an experimental database, with a machine learning model that\nuses carefully curated and physics-inspired features derived from\nX-ray crystal structures to predict backbone chemical shifts for proteins.\nIn this work, we extend the UCBShift 1.0 method to side chain chemical\nshift prediction to perform whole protein analysis, which, when validated\nagainst well-defined test data shows higher accuracy and better reliability\ncompared to the popular SHIFTX2 method. With the greater abundance\nof cleaned protein shift-structure data and the modularity of the\ngeneral UCBShift algorithms, users can gain insight into different\nfeatures important for residue-specific stabilizing interactions for\nprotein backbone and side chain chemical shift prediction. We suggest\nseveral backward and forward applications of UCBShift 2.0 that can\nhelp validate AlphaFold structures and probe protein dynamics.