Figure 4.

Performances of different tools on human data set B. LncFinder had the best accuracy of 0.9728. CPC had a strong tendency to classify lncRNAs as protein-coding transcripts and thus having low accuracy of 0.8304 (web server). As an upgraded version, CPC2 presented accuracy of 0.9614. CPC2 was a big improvement on its predecessor and also outperformed CNCI and PLEK that obtained accuracies of 0.9450 and 0.9274, respectively. CPAT (re-trained model) was inferior to only LncFinder and obtained an accuracy of 0.9642. Even when secondary structure-derived features were excluded, LncFinder (Without.SS) can still surpass other tools with an accuracy of 0.9716.