Pairs of plots showing the heavy-atom RMSDs (left) and sensitivities (right) as a function of radius of gyration and total number of base-base contacts of all decoy structures for (a) yeast tRNA^Phe (b) the Alu domain of the Mammalian SRP (c) the hammerhead ribozyme, (d) the HCV IRES (e) the P4-P6 domain of the Tetrahymena group I intron folded with traditional tertiary restraints, and (f) the P4-P6 domain of the Tetrahymena group I intron folded with traditional and MOHCA tertiary restraints. Results from simulations of the crystal structures are marked with asterisks. If two or more resulting structures had both their total contact numbers and their radii of gyration corresponding to the same region in the plot, their RMSDs and sensitivities were averaged. White indicates regions where there were no decoys with a combination of contact number and radius of gyration. The tendency for the best structures to have both a high number of contacts and a large radius of gyration is somewhat less pronounced for the hammerhead ribozyme, but the best structure still had the highest score given the scoring function. This is consistent with the results of running five independent simulations of the crystal structures of the tRNA^Phe, SRP Alu domain, the HCV IRES and the P4-P6 domain of the group I intron. The radii of gyration, total numbers of contacts, RMSDs, and sensitivities of each were averaged, plotted along with the decoys and marked with an asterisk. The RMSD plot for the HCV IRES does not show a correlation due to the alternate accessible conformation seen in the predicted structure. However, its sensitivity plot does show a trend for high sensitivity structures to have a high R_g and a high number of contacts.

The hepatitis C virus internal ribosomal entry site. Panel (a) shows the crystal structure of the molecule (PDB 1KH6), (b) is the predicted structure of the molecule, and (c) is an overlay of the crystal structure (red) and the predicted structure (blue). Bases that were disordered or that were involved in dimerization in the crystal structure were omitted from the figure and from the calculation of the RMSD, sensitivity, and PPV, even though they were present in the simulations.

Sacchromyces cerevisiae tRNA^Phe. Panel (a) shows the crystal structure of the molecule (PDB 1EHZ), (b) is the predicted structure of the molecule, and (c) is an overlay of the crystal structure (red) and the predicted structure (blue).

P4-P6 domain of the Tetrahymena group I intron, supplemented with MOHCA data. Panel (a) shows the crystal structure of the molecule (PDB 1GID), (b) is the predicted structure of the molecule, and (c) is an overlay of the crystal structure (red) and the predicted structure (blue).

Plots comparing the van der Waals potential energy for the usual 6–12 Lennard-Jones potential at 10% strength and the soft core potential with λ = 0.1 for guanine and cytosine bases interacting with an aromatic carbon probe atom. (a) A guanine base with the 6–12 potential at 10% strength. (b) A guanine base with the soft core potential with λ = 0.1. (c) A cytosine base with the 6–12 potential at 10% strength. (d) A cytosine base with the soft core potential with λ = 0.1. Parameters are taken from the AMBER ff99 force field relative to an aromatic carbon probe atom like the ones that make up nucleobases. Adenine and uracil plots are available in supplementary material. In all cases, the 6–12 potential has a local minimum at the center of each ring with barriers of escape in excess of 1 million kcal/mol, leaving the chance that two rings could interlock inseparably in an unphysical conformation when the potential is first restored. In contrast, the soft core potential has a local maximum at the center of the bases with no local energy minima inside any rings, and the energies inside each are about six orders of magnitude lower. Bases that were overlapping or interlocked during the stage of the run without VDW or electrostatic forces will smoothly slide apart.

The Alu domain of the mammalian signal recognition particle. Panel (a) shows the crystal structure (PDB 1E8O), (b) is the predicted structure of the molecule, and (c) is an overlay of the crystal structure (red) and the predicted structure (blue). Structure images were generated using PyMOL.⁴⁴

The full-length hammerhead ribozyme. Panel (a) shows the crystal structure of the molecule (PDB 2GOZ), (b) is the predicted structure of the molecule, and (c) is an overlay of the crystal structure (red) and the predicted structure (blue).

P4-P6 domain of the Tetrahymena group I intron. Panel (a) shows the crystal structure of the molecule (PDB 1GID), (b) is the predicted structure of the molecule, and (c) is an overlay of the crystal structure (red) and the predicted structure (blue).