Fig. 5

Evaluating the performance of different methods in ORF prediction. a An illustration of how the frame test was performed to predict ORFs from rRibo-seq data by Ribo-TISH. b ROC curves across six strategies of ORF predictions implemented in Ribo-TISH, RiboTaper, and ORF-RATER. An RPKM value of 1 was used as a cutoff to define actively translating genes for positive and negative sets. c The short ORFs (<100 aa) of CCDS in Ensembl or d the experimentally validated uORFs curated by uORFdb were used as a positive set (RPKM ≥ 1) for the ROC analyses across 7 strategies implemented in Ribo-TISH, RiboTaper, ORF-RATER, and riboHMM. ROC curves across six strategies implemented in Ribo-TISH, RiboTaper, and ORF-RATER when the annotated ORFs of CCDS in Ensembl, with RPKM e between 0.1 and 0.5 or f between 0.5 and 1, are used as a positive set, respectively