Fig. 1

a Five steps were performed to predict novel lncRNAs. The X-axis stood for each step, and Y-axis stood for the number of transcripts left after each step. First, single-exon transcripts were filtered out. Second, transcripts longer than 200 nt were kept. Third, using Cuffcompare, transcripts overlapped with exons of annotated mRNAs were assigned as annotated lncRNAs. Fourth, transcripts expressed at low levels (FPKM < 0.5) were removed. Fifth, four programs were applied to predict the coding potential of transcripts, namely CNCI, CPC, Pfam-scan, and phyloCSF. b The Venn diagram showed numbers of predicted non-coding transcripts by CNCI, CPC, Pfam-scan, and phyloCSF. A total of 1147 transcripts predicted as non-coding transcripts by all the algorithms were characterized as novel lncRNAs