Table 2.

Confusion matrix for most divergent calibrated sets

For each Rfam family, the classification of each RNA sequence from their actual class (y-axis) into their predicted classes (x-axis) is shown. Q^M is the specificity of the method and Q^D the sensitivity (see Materials and Methods section for definition). This confusion matrix was calculated from training and testing sets that were at most 20% similar and used calibrated graph properties. We observed that the chosen graph properties calculated for the chosen graph representation are best able to reflect the Rfam families SECIS, Intron gp II, 5S rRNA, tRNA, MIRNA and HACA-BOX. Entries are coloured by number classified starting from 0 (white) to 2000 (green).