Unifying evolutionary and thermodynamic information for RNA folding of multiple alignments

Seemann, Stefan E.; Gorodkin, Jan; Backofen, Rolf

where P t is the number of mutual base pairs in the two assignments (true positives), N t the number of mutual pairs of bases not base pairing (true negatives), P f the number of predicted base pairs not in the Rfam assignment (false positives) and N f the number of base pairs in the Rfam assignment not predicted to pair (false negatives). We also calculated sensitivity, SEN ¼ P t =ðP t þ N f Þ and positive predictive value, PPV ¼ P t =ðP t þ P f Þ. Their geometric mean is a good approximation to the MCC for base pair prediction (43) . The parameter tuning was done by optimizing the MCC to get a best possible tradeoff between sensitivity and PPV. In addition, we used the more stringent structure evaluation scheme R 5 introduced by Gorodkin et al.