RibORF: Identifying Genome-Wide Translated Open Reading Frames Using Ribosome Profiling

Abstract

Ribosome profiling identifies RNA fragments associated with translating ribosomes. The technology provides an opportunity to examine genome-wide translation events at single-nucleotide resolution and in an unbiased manner. Here I present a computational pipeline named RibORF to systematically identify translated open reading frames (ORFs), based on read distribution features representing active translation, including 3-nt periodicity and uniformness across codons. Analyses using the computational tool revealed pervasive translation in putative 'noncoding' regions, such as lncRNAs, pseudogenes, and 5'UTRs. The computational tool is useful for studying functional roles of non-canonical translation events in various biological processes. © 2018 by John Wiley & Sons, Inc.

Keywords: noncoding RNA; open reading frame; ribosome profiling; translation.

Figure 1

For candidate ORFs with ribosome profiling reads, the read distribution features can separate them into three groups, including actively translated ORFs, off-frame ORFs, and non-ribosomal protein-RNA complex binding.

Figure 2

The outline of the RibORF algorithm.

Figure 3

Types of candidate ORFs, defined based on transcript types and ORF locations in reference transcripts.

Figure 4

Check read distribution around the start and stop codons of canonical ORFs, and correct read locations based on offset distances between 5′ end of fragments and ribosomal A-sites. (A) Ribosome profiling reads were grouped based on fragment length (i.e. 28nt, 29nt and 30nt). The plots show the distribution of 5′ end of read fragments around the start and stop codons of canonical ORFs of mRNAs. The dataset shows high quality and clear 3-nt periodicity. The summary statistics of read fragments were shown in the box. (B) An example of low quality ribosome profiling dataset, which does not show obvious 3-nt periodicity. (C) Distribution of offset corrected read distribution around the start and stop codons of canonical ORFs.

Figure 5

Example outputs of RibORF program and algorithm performance evaluation. (A) Example candidate ORFs with training parameters and predicted translated P-values. These values were shown in the files “pred.pvalue.parameters.txt” and “repre.valid.pred.pvalue.parameters.txt”. (B) The predicted translated P-value cutoffs and associated statistics of true positive, false positive, true negative, false negative, false positive rate and true positive rate, as shown in the “stat.cutoff.txt” file. (C) The ROC curve showing the performance of the RibORF program in identifying translated ORFs. The plot was shown in the “plot.ROC.curve.pdf”. The AUC value was shown in the plot.