Monday, January 25, 2010

RMSD is not the only metric in the world.

RNA, Vol 15, 1875-1885, 2009
Parisien and Major hit another big methodology "goooool (read it in Spanish)" in the RNA bioinformatics little world.

They propose to use two new quantities, a deformation index, and a deformation profile, to better take into account the local similarity of nucleotides based on the stacking and base-pairing interaction geometry. This in contrast to the more common backbone-centric view, or the all-atom view, which traditionally has stuck to the RMSD metric.

The deformation index is just the ratio of the RMSD to the Mathews Correlation Coefficient. The Deformation Profile is a square matrix whose dimensions are the length of the RNA sequence, and it's values are the average inter-atomic distances between the pyridine ring of bases from a predicted structure with respect to a reference structure.

It has always been interesting to me how Major's group ideas have always felt to me as having a chemical graph theory flavour, which clearly comes from the computer scientist world. For some reason which eludes me, it's still to this day more trendy to approach the RNA bioinfo world from the C.S. side, that the chem. graph theory side. I should try and use this marketing trick myself.