A Support Vector Machine based method to distinguish long non-coding RNAs from protein coding transcripts

BMC Genomics

Table 3 Results for models trained with human data

Test data set
Method	GRCh37	GRCh38	NONCODE
Radial using GRCh37 and first ORF
Sensitivity	9 8 . 9 5 %	9 9 . 4 3 %	*96.67%*
Specificity	97.41%	97.23%	-
Accuracy	9 8 . 1 8 %	98.33%	-
Radial using GRCh38 and first ORF
Sensitivity	89.86%	97.54%	88.75%
Specificity	9 8 . 6 4 %	9 9 . 2 6 %	-
Accuracy	94.25%	9 8 . 4 0 %	-
CPC^a,e
Sensitivity	67.23%	69.90%	-
Specificity	97.62%	73.90%	-
Accuracy	82.43%	71.90%	-
CPAT^a,e
Sensitivity	94.60%	89.90%	-
Specificity	85.28%	92.40%	-
Accuracy	89.94%	91.20%	-
lncRScan-SVM^a
Sensitivity	93.88%	-	-
Specificity	89.20%	-	-
Accuracy	91.94%	-	-
iSeeRNA^b,c
Sensitivity	96.10%	-	-
Specificity	94.70%	-	-
Accuracy	95.40%	-	-
lncRNApred^d,f
Sensitivity	-	-	93.40%
Specificity	-	-	-
Accuracy	-	-	-
FEELnc^e
Sensitivity	-	92.30%	-
Specificity	-	91.50%	-
Accuracy	-	91.90%	-

Results in bold are the best for each test data set. Note that our method produced the best results
^aResults obtained in Han et al. [25]
^bResults obtained in Sun et al. [27]
^cThis method was created to classify only lincRNAs
^dResults obtained in Sun et al. [24]
^eResults obtained in Wucher et al. [28]
^fWe only considered sensitivity, since the negative test data was not clearly specified in the article

ISSN: 1471-2164