Two aspects of ribonucleic acids (RNA) three-dimensional (3-D) structures which determine the feasibility of a modeling problem and the precision of the models were studied: A. the conformational space of nucleotides and B. RNA structural data. A.) The conformational space of RNA is determined by the conformational space of nucleotides. A database of nucleotide conformations taken from RNA structures determined from Nuclear Magnetic Reasonnance (NMR) and crystallographic data was constituted. Compared to the consensus models and for nucleotide conformational sets of the same length, the ones derived from a classification by root-mean-square (rms) deviation between all atoms of each nucleotide pair reproduce RNA hairpin loops with a better precision. The number of nucleotide conformations necessary to reproduce these hairpin loops is small compared to the number theoretically allowed by the seven free torsion angles. Although small, conformational sets that produce good precision models introduce such a flexibility in the RNA chain that spatial overlappings of atoms occur. In order to eliminate such stereochemical inconsistent models and keep the number of models manageable (especially for large RNA), steric constraints must be added. B.) Although the number of conformations used for each nucleotide determines the size of the conformational space, the available structural data is the most influencing factor on the algorithm performances, and therefore determines the feasibility of a problem. The modeling of transfer RNA (tRNA) indicates that the knowledge of one base-pairing coupled with the secondary structure, inferred from phylogeny analysis or empirical programs, is necessary and sufficient to infer tRNA tertiary folds. Other structural information was used to reduce the loops conformational space. All tRNA structural information used to produce the models which are of acceptable precision (all atoms rms deviations between 3 and 4A to the crystal structure) was available before the crystal model was determined, indicating the prediction possibilities of the procedure. The modeling of other large RNA is in progress.
