Mathews correlation coefficient (MCC) of different ways to score the
conservation term in RNAalifold, on the CMfinder SARSE data-set.
Best (if applicable) is the best performance achieved, the other uses the (old) default parameters.
All computations (except the 2002 variant) were done using the new variant of
not counting gaps as bases for energy computations.
| Computation type |
MCC |
| RIBOSUM best |
0.936 |
| New best |
0.916 |
| 2002 (Old) best |
0.884 |
| RIBOSUM |
0.878 |
| Pfold-like best |
0.876 |
| New |
0.845 |
| 2002 |
0.831 |
| weighted |
0.828 |
| Pfold-like |
0.760 |
Mean results when using the parameters
of Andronescu
et al. versus the usual Turner parameters. While some predictions are
better when Andronescu's parameters are used, on average they do slightly
worse than the Turner parameters.
| Computation type |
Dataset |
MCC Turner |
MCC Andronescu |
| RIBOSUM |
CMfinder |
0.936 |
0.904 |
| New |
CMfinder |
0.845 |
0.832 |
| 2002 |
CMfinder |
0.831 |
0.819 |
| RIBOSUM |
RNASTRAND-Rfam |
0.759 |
0.758 |
| New |
RNASTRAND-Rfam |
0.710 |
0.688 |
| RIBOSUM |
Rfam |
0.790 |
0.766 |
| New |
Rfam |
0.746 |
0.727 |
| 2002 |
Rfam |
0.729 |
0.718 |