Automated de novo prediction of native-like RNA tertiary structures

Abstract

RNA tertiary structure prediction has been based almost entirely on base-pairing constraints derived from phylogenetic covariation analysis. We describe here a complementary approach, inspired by the Rosetta low-resolution protein structure prediction method, that seeks the lowest energy tertiary structure for a given RNA sequence without using evolutionary information. In a benchmark test of 20 RNA sequences with known structure and lengths of approximately 30 nt, the new method reproduces better than 90% of Watson-Crick base pairs, comparable with the accuracy of secondary structure prediction methods. In more than half the cases, at least one of the top five models agrees with the native structure to better than 4 A rmsd over the backbone. Most importantly, the method recapitulates more than one-third of non-Watson-Crick base pairs seen in the native structures. Tandem stacks of "sheared" base pairs, base triplets, and pseudoknots are among the noncanonical features reproduced in the models. In the cases in which none of the top five models were native-like, higher energy conformations similar to the native structures are still sampled frequently but not assigned low energies. These results suggest that modest improvements in the energy function, together with the incorporation of information from phylogenetic covariance, may allow confident and accurate structure prediction for larger and more complex RNA chains.

Fig. 1.

A simple energy function for RNA fragment assembly. (A) Coordinate system set up on one base to define the potential. (B) Distribution of Δx, Δy coordinates for a uridine residue near adenosine residues in the ribosome crystal structure, smoothed with a 2-Å Gaussian filter. The logarithm of this distribution provides a knowledge-based potential for de novo RNA structure prediction. Interactions with the three different edges of adenosine (Watson–Crick, Hoogsteen, and sugar) correspond to positions in the three sectors of the map demarcated by the dotted lines and the negative x axis. (C and D) Distributions of angle between base planes (C) and relative stagger of base planes (D) for base pairs observed in the large ribosome crystal structure (blue) and models produced without and with coplanarity terms (gray and red, respectively).

Fig. 2.

Finding the native structure of a small RNA hairpin by FARNA. (A) Histograms of rmsd (calculated over C4′ atoms) between models generated by 5,000 cycles of Monte Carlo fragment assembly without the influence of any energy terms (gray) and with successive addition of the following terms to the energy function: radius-of-gyration (cyan); steric penalties (magenta); Watson–Crick-edge component of the base-pairing term (blue); Hoogsteen and sugar-edge components of base pairing and base stacking (green); and coplanarity terms (red). (B) Native structure of the hairpin [first model from NMR ensemble 1ZIH (26)]. (C–E) Lowest energy structures from simulations with radius-of-gyration term and steric penalties (C), plus Watson–Crick-edge component of the base-pairing term (D), and the full energy function (E). The residues G5 and A8, which form a sheared base pair in the native structure, are highlighted. In this figure and following figures, the coloring scheme shows rainbow coloring for the backbone (cartoons); and adenosine, cytidine, guanosine, and uridine bases are orange, green, blue, and red, respectively. Residues discussed in Results are rendered with thicker lines. Figures of molecules prepared in Pymol (Delano Scientific).

Fig. 3.

Best of five FARNA cluster centers (Left in each panel) and native structures (Right in each panel) for a curved RNA helix incorporating an internal loop with G-A and A-A non-Watson–Crick base pairing [283D (31)] (A); stem loop SL2 of the HIV-1 PSI RNA packaging signal [1ESY (34)] (B); domain 5 from the Pyaiella littoralis group II intron [2F88 (35)] (C); and the frameshifting RNA pseudoknot from beet western yellow virus [1L2X (39)] (D). (A and B) Magnified superpositions of noncanonical base pairs (native in white; model in color) are displayed (Lower).

Fig. 4.

Best of five FARNA cluster centers (Left), native structure (Center), and model with best recovery of non-Watson–Crick base pairs from overall FARNA population (Right) for the HIV-1 Rev response element high-affinity site [1CSL (52)]. Residues discussed in Results are rendered with thicker lines.