Utilizing Machine Learning to Accelerate Automated Assignment of Backbone NMR Data

Nuclear magnetic resonance (NMR) spectroscopy is a powerful method for determining three-dimensional structures of biomolecules, including proteins. The protein structure determination process requires measured NMR values to be assigned to specific amino acids in the primary protein sequence. Unfortunately, current manual techniques for the assignment of NMR data are time-consuming and susceptible to error. Many algorithms have been developed to automate the process, with various strengths and weaknesses. The algorithm described in this paper addresses the challenges of previous programs by utilizing machine learning to predict amino acid type, thereby increasing assignment speed. The program also generates place-holders to accommodate missing data and amino acids with unique chemical characteristics, namely proline. Through machine learning and residue-type tagging, the assignment process is greatly sped up, while maintaining high accuracy.