Comparison of Variants of Uncertain Significance in Three Regions of the Human Glucokinase Protein Using In Vitro and In Silico Analyses.

Abstract

There is a growing field of research focusing on the bioinformatic analysis of human genetic variation and the associated diseases. To study how well in vitro testing of purified proteins compares to bioinformatic variant prediction, we chose to analyze glucokinase (GCK) missense variations between residues 119-132, 257-262, and 412-427. These regions contained a large number of variants of uncertain significance (VUS) as well as a few pathogenic variants to use for comparison. We compared experimentally produced Vmax values from purified GCK variant proteins to predictive methods such as molecular dynamics simulation, ConSurf, iStable, the evolutionary model of variant effect (EVE), PredictSNP, and calculated binding energy. After determining which variants are pathogenic or benign based on experimental results or previous genetic studies, we found that ConSurf was the best at predicting pathogenicity. Interestingly, one VUS, D262N, showed an increase in activity and thus was difficult to interpret as pathogenic or benign. This study is an attempt to provide a framework for the utility of missense variant predictive programs.