Evaluating protein structure descriptors and tuning Gauss integral based descriptors

The general development in the natural sciences from relative comparison to absolute description forces the current relative protein structure comparison, based on similarity measures, to be supplemented, or even replaced, by absolute description of each individual protein structure. This paper addresses the question of what should be required from a good set of protein structure descriptors. As an example a Gauss integral based family of protein structure descriptors, that has been shown to successfully classify the geometry of CATH2.4 connected protein domains, is examined. The CATH2.4 domains are here observed to break a symmetry under reversal of the direction of traversal of the protein backbone that general folded tubes possess. It is thus a challenge for any large scale protein or polymer model to explain this broken symmetry.