Long Noncoding RNA Identification: Comparing Machine Learning Based Tools for Long Noncoding Transcripts Discrimination

Table 3

Overview of each tool’s performance on different testing datasets.

Testing dataset	CPC	CPAT	CNCI	PLEK	LncRNA-ID	lncRScan-SVM
Human MCF-7 (PacBio)^1
Specificity	99.90		91.80	94.70
Sensitivity	19.00		78.70	95.80
Accuracy	97.00		91.30	94.70
Human HelaS3 (454)^2
Specificity	99.90		93.90	95.50
Sensitivity	47.20		81.10	92.50
Accuracy	99.00		93.70	95.40
Human (from GENCODE)^3
Specificity	99.97	99.55		89.18	95.28
Sensitivity	66.48	86.95		99.52	96.28
Accuracy	83.22	93.25		94.32	95.78
Mouse (from GENCODE)^4
Specificity	98.75	98.95		70.94	92.10
Sensitivity	76.55	38.80		88.11	94.45
Accuracy	87.65	68.88		79.49	93.28
Human (from GRCh37/hg19)^5
Specificity	97.62	85.28				89.20
Sensitivity	67.23	94.60				93.88
Accuracy	82.43	89.94				91.94
Mouse (from GRCm38/mm10)^5
Specificity	98.37	88.17				89.14
Sensitivity	75.46	95.34				95.29
Accuracy	86.91	91.76				92.21

The results of the tools being tested on the same datasets are listed above. Bold numbers denote the highest value of the metrics.
¹MCF-7 is available at http://www.pacb.com/blog/data-release-human-mcf-7-transcriptome/; ²dataset of HelaS3 is available at https://www.ncbi.nlm.nih.gov/sra/SRX214365; ^3,4datasets are available at https://www.dropbox.com/sh/7yvmqknartttm6k/AAAQHvLZPjgjf4dtmHM7GNCqa/H1_gencode?dl=0 and https://www.dropbox.com/sh/7yvmqknartttm6k/AACzaG-QJggvbXW6LA32oo7ba/M1_gencode?dl=0; ⁵dataset of human and mouse is available at http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0139654.