Statistical analysis of protein ensembles

As 3D protein-configuration data is piling up, there is an ever-increasing need for well-defined, mathematically rigorous analysis approaches, especially that the vast majority of the currently available methods rely heavily on heuristics. We propose an analysis framework which stems from topology, the field of mathematics which studies properties preserved under continuous deformations. First, we calculate a barcode representation of the molecules employing computational topology algorithms. Bars in this barcode represent different topological features. Molecules are compared through their barcodes by statistically determining the difference in the set of their topological features. As a proof-of-principle application, we analyze a dataset compiled of ensembles of different proteins, obtained from the Ensemble Protein Database. We demonstrate that our approach correctly detects the different protein groupings.