Skip to main content

Table 2 Secondary structure prediction accuracy with and without speed improvements for tRNA and 5S rRNA sequences.

From: Detection of non-coding RNAs on the basis of predicted secondary structure formation free energy change

  Without speed improvements With speed improvements
RNA type Sensitivity PPV Best sensitivity Sensitivity PPV Best sensitivity
tRNA 92.9 ± 12.6 92.7 ± 14.3 98.8 ± 4.3 93.0 ± 12.3 92.7 ± 14.1 99.1 ± 2.8
5S rRNA 91.7 ± 7.0 82.0 ± 7.1 97.9 ± 3.2 91.7 ± 6.9 81.9 ± 7.0 98.0 ± 3.0
  1. Sensitivity is the percent of known base pairs correctly predicted. Positive predictive value (PPV) is the percent of predicted base pairs that are in the known structure. Best sensitivity is the sensitivity of the most sensitive structure in a set of 750 suboptimal structures, i.e. structures with folding free energy change similar to the lowest free energy structure. Sensitivity and positive predictive value are calculated as described previously [46]. The tRNA and 5S rRNA datasets are sets of randomly chosen sequences, set sizes 40 and 14, respectively, used for benchmarks previously [46]. Accuracies are reported as averages for all pairwise combinations of sequences, and single standard deviations are reported as errors. The average accuracy of secondary structure prediction by Dynalign is essentially unchanged by pre-filtering base pairs and changing the implementation of the M parameter. Without speed improvements, M was set to 15 for both tRNA and 5S rRNA. With speed improvements, M was 6 for tRNA and 7 for 5S rRNA for these benchmarks.