Follow the allosteric transitions to predict variant pathogenicity : a channel-specific approach

We were intrigued to read a recent article by Brünger et al. outlining an innovative method to predict pathogenicity for variants in ion channel-encoding genes by determining the 2D distance of amino acids from the central axis of the ion channel pore using published high-resolution protein structures. The resulting dataset was combined with amino acid properties and then machine learning algorithms to describe the potential pathogenicity of genetic variants. A major advance of this approach is the clear distinction in the distance to the pore axis for pathogenic versus benign population variants: the closer to the pore axis, the greater the likelihood of pathogenicity. This enabled correlation with clinical representations whereby more severe phenotypes were observed for variants closest to the pore. We find this an elegant approach to a complex problem. However, the universal approach ultimately limits precision in predicting pathogenicity, which may be overcome with a protein-specific approach.