Secondary structural entropy in RNA switch (Riboswitch) identification

BMC Bioinformatics

Table 6 Classification performance using cross validation

Classifier	TP rate	FP rate	MCC	R.O.C. area
{L,GC,GU,Sil}	0.750	0.250	0.500	0.826
{L,GC,GU}	0.673	0.327	0.346	0.700
{L,GC,GU,BJK}	0.644	0.356	0.289	0.691
{L,GC,GU,BJKbp}	0.654	0.346	0.309	0.690
{L,GC,GU,RND}	0.654	0.346	0.308	0.689
{L,MFE,GC,GU,RND}	0.673	0.327	0.346	0.714
{L,MFE,GC,GU}	0.654	0.346	0.308	0.707
{L,MFE,GC,GU,BJK}	0.663	0.337	0.327	0.703
{L,MFE,GC,GU,BJKbp}	0.663	0.337	0.327	0.701
{L,MFE,GC,GU,Sil}	0.625	0.375	0.250	0.697
{L,MFE,GU}	0.663	0.337	0.327	0.710
{L,MFE,GU,RND}	0.663	0.337	0.327	0.702
{L,MFE,GU,BJKbp}	0.663	0.337	0.32	0.701
{L,MFE,GU,Sil}	0.654	0.346	0.308	0.701
{L,MFE,GU,BJK}	0.644	0.356	0.289	0.699
{L,MFE,GC,RND}	0.663	0.337	0.327	0.708
{L,MFE,GC,BJK}	0.663	0.337	0.327	0.703
{L,MFE,GC,BJKbp}	0.635	0.365	0.269	0.683
{L,MFE,GC}	0.606	0.394	0.212	0.650
{L,MFE,GC,Sil}	0.635	0.365	0.270	0.644
{L,MFE,GCU,RND}	0.644	0.356	0.289	0.693
{L,MFE,GCU,BJK}	0.625	0.375	0.250	0.617
{L,MFE,GCU,BJKbp}	0.596	0.404	0.193	0.595
{L,MFE,GCU}	0.587	0.413	0.174	0.581
{L,MFE,GCU,Sil}	0.548	0.452	0.097	0.554

104-fold binomial logistic classifiers on all of the 52 riboswitch sequences and their antisense sequences. classifier features shown in legend. WekaⒸopen source software package used. Features L,MFE,GC,GU,GCU and U denote length, MFE, GC-composition, and uracil frequency, respectively. Features RND and BJK denote structural entropy of the RND and BJK models, respectively. as defined in [54]. Feature BJKbp denotes base-pairing entropy as defined in [54]. Feature Sil denotes the two-cluster average Silhouette index of energy landscape as calculated in [46].

ISSN: 1471-2105