Table 1.
Accuracy of RNA secondary structure models.
| RNA | Length★ | Number of helices†
|
||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Cryst. | No data
|
1D
|
1D + 2D
|
2D
|
||||||
| TP | FP | TP | FP | TP | FP | TP | FP | |||
| Adenine riboswitch‡ | 71 | 3 | 2 | 3 | 3 | 0 (1) | 3 | 0 (1) | 3 | 0 (1) |
| tRNAphe | 76 | 4 | 2 | 3 | 3 | 1 | 4 | 0 | 4 | 0 |
| P4–P6 RNA | 158 | 11 | 10 | 1 | 9 | 2 | 9 | 2 | 11 | 0 |
| 5S rRNA | 118 | 7 | 1 | 9 | 6 | 3 | 7 | 0 (1) | 7 | 0 (1) |
| c-di-GMP riboswitch‡ | 80 | 8 | 6 | 2 | 6 | 2 | 7 | 1 | 7 | 1 |
| Glycine riboswitch‡ | 158 | 9 | 5 | 3 | 8 | 1 | 9 | 0 | 9 | 0 |
| Total | 661 | 42 | 26 | 21 | 35 | 9 (10) | 39 | 3 (5) | 41 | 1 (3) |
| False negative rate§ | 38.1% | 16.7% | 7.1% | 2.4% | ||||||
| False discovery rate|| | 44.7% | 20.4 (22.2)% | 7.1 (11.4)% | 2.3 (6.8)% | ||||||
Length of RNA in nucleotides.
Cryst, number of helices in crystallographic model; TP, true positive helices; FP, false positive helices; 1D, models using one-dimensional SHAPE chemical mapping data; 2D, models using mutate-and-map data. For FP, a helix was considered incorrect if its base pairs did not match the majority of base pairs in a crystallographic helix. Numbers in parentheses required that the matching crystallographic base pairs have Watson–Crick geometry.
Ligand-binding riboswitches were probed in the presence of small-molecule partners (5 mM adenine, 10 μM cyclic di-guanosine-monophosphate or 10 mM glycine). All experiments were carried out with 10 mM MgCl2, 50 mM Na-HEPES, pH 8.0.
False negative rate = (Cryst–TP)/TP.
False discovery rate = FP/(FP + TP). Numbers in parentheses count matches of model base pairs to non-Watson–Crick crystallographic base pairs as false discoveries.