We present a universal nucleic acids structural alphabet suitable for conformational analysis of DNA as well as RNA structures. The underlying local conformational classes assigned at a level of dinucleotide “step” are characterized by seven torsion angles of the sugar-phosphate backbone and two torsion angles around two glycosidic bonds, supplemented by empirically determined parameters. Our previous definition of 44 DNA conformational classes [1,2] was extended to incorporate also RNA conformers.
The conformational classes are built on a subset of RNA conformers defined previously [3] and classes newly defined by means of hierarchical clustering of available structural data. The necessary structural information was extracted from a sequentially non-redundant set of high resolution crystal structures containing about 60 and 60 thousands steps for DNA and RNA, respectively. We have obtained more than 70 conformational classes covering the combined structural variability of DNA and RNA.
We have found that many of the previously defined DNA conformational classes are shared among nucleic acids and are suitable for RNA description. They differ in relative occurrence, whereas in DNA B-forms prevail and A-forms are less common, A-forms are the most prevalent in RNA. The most significant feature of new conformational classes unique for RNA is the high occurrence of non-stacked steps. In spite of the apparent structural flexibility of RNA, the conformations rarely found in DNA comprise only about one third of all characterized classes.