Bayesian multilevel models for assessing the quality of NMR resolved protein structures

ARROYUELO, AGUSTINA.; MARTÍN, OSVALDO ANTONIO

Montevideo

Congreso; KHIPU Latin American Meeting in Artificial Intelligence; 2019

We present a framework based on a multilevel Bayesian model for NMR resolved protein structure validation, through corrected 13Cα chemical shifts. Said framework provides a hierarchical regression model between experimental and theoretical  13Cα where partial pooling is applied  based on aminoacid identity. By computing the posterior predictive density of this model we obtain corrected theoretical 13Cα chemical shifts, and use  the differences between observed and corrected values for protein structure validation.