Results based on alternative datasets (database alignments as well)
datasets taken from
ILM
[ Ruan et al.,
2003 ]
The hxmatch predictions were calculated using hxmatch -A with default
parameters. The "-A" option means, that the RNAalifold (see
also Vienna RNA package,
v1.5beta version)
prediction is included in the computation of the initial weight matrix.
Raw corresponds to the consensus structures as predicted by
hxmatch -A, filled are the
structures obtained by computing the thermodynamically most favorable
structure consistent with the consensus structure (using
RNAfold
-C ).
Nseq gives the number of sequences contained in the alignment;
eta is the mean pairwise sequence identity of the alignment.
PK =
(number of correctly predicted pseudoknots)/(number of pseudoknots in the
reference structure);
SS = 100*TP/RP and SP = 100*TP/(TP+FP);
RP = number of
base pairs in
the reference structure;
TP = number of true positive
predicted base
pairs;
FP = number of false positive predicted base pairs;
Compared gives the predicted structure compared to the
reference structure in linked diagram representation.
Base pairs drawn in black are present in both the
predicted structure
and the reference structure (true positives). Red base pairs are only
present in our prediction, but not in the reference (false
positives). And
base pairs which are not contained in the prediction, but are part of
the reference structure (false negatives) are drawn in green.
The slight difference of SS and SP of ilm compared to the values given in
Ruan et al., 2003
is due to the use of a later version of ilm.
Last modified: 2004-09-02 16:53:34 xtina