Combining in silico prediction and ribosome profiling in a genome-wide search for novel putatively coding sORFs

Abstract

BackgroundIt was long assumed that proteins are at least 100 amino acids (AAs) long. Moreover, the detection of short translation products (e.g. coded from small Open Reading Frames, sORFs) is very difficult as the short length makes it hard to distinguish true coding ORFs from ORFs occurring by chance. Nevertheless, over the past few years many such non-canonical genes (with ORFs < 100 AAs) have been discovered in different organisms like Arabidopsis thaliana, Saccharomyces cerevisiae, and Drosophila melanogaster. Thanks to advances in sequencing, bioinformatics and computing power, it is now possible to scan the genome in unprecedented scrutiny, for example in a search of this type of small ORFs.ResultsUsing bioinformatics methods, we performed a systematic search for putatively functional sORFs in the Mus musculus genome. A genome-wide scan detected all sORFs which were subsequently analyzed for their coding potential, based on evolutionary conservation at the AA level, and ranked using a Support Vector Machine (SVM) learning model. The ranked sORFs are finally overlapped with ribosome profiling data, hinting to sORF translation. All candidates are visually inspected using an in-house developed genome browser. In this way dozens of highly conserved sORFs, targeted by ribosomes were identified in the mouse genome, putatively encoding micropeptides.ConclusionOur combined genome-wide approach leads to the prediction of a comprehensive but manageable set of putatively coding sORFs, a very important first step towards the identification of a new class of bioactive peptides, called micropeptides.

Decoding sORF translation – from small proteins to gene regulation

“…New translated regions have not only been identified in transcripts thought to be non-coding, but also upstream of a large fraction of protein-coding ORFs (so-called upstream ORFs, uORFs). 4,[13][14][15][16][17][18][19][20][21] These usually short translated ORFs were likely missed in previous mutagenesis screens due to their small size, and remained un-annotated in genome annotations due to their small size and lack of evidence for codingness. [22][23][24] In analogy to the term 'pervasive transcription', 25 this unforeseen prevalence of short ORF (sORF) translation has spurred the notion of 'pervasive translation'.…”

New Peptides Under the s(ORF)ace of the Genome

“…Translation inhibitors, such as harringtonine (HR), which generates a pile-up of RBFs at the start codon, have also been used to identify translation initiation sites in actively translated ORFs [30]. Some of these studies have focused exclusively on the identification of translated smORFs in fruit flies [32], zebrafish [31], yeast [36], and mice [41].…”

Assessing the Performances of Protein Function Prediction Algorithms from the Perspectives of Identification Accuracy and False Discovery Rate

“…As demonstrated in Table 2 and Supplementary Table S2, the genome scanning revealed that the number of proteins in any of those studied 93 families predicted by SVM, PNN and KNN did not exceed 10% of the total number of proteins in the whole genome, and this was the same situation for the majority (82%) of the studied 93 families by BLAST. The higher number of proteins predicted for certain functional family may indicate a higher false discovery rate [113,114]. For human genome, the number of proteins identified by SVM was equivalent to or was slightly higher than that of both PNN and KNN, but was significantly lower than that of BLAST (Figure 2a).…”

“…As reported, genome scanning was a comprehensive method to evaluate the capacity of protein functional prediction tools in identifying and classifying protein family [113,114]. In this paper, an evaluation on the false discovery rate of the studied protein function predictors was performed by scanning the genomes of 4 model organisms representing 4 species kingdoms (homo sapiens from Animalia, arabidopsis thaliana from Plantae, saccharomyces cerevisiae from Fungi and mycobacterium tuberculosis from Bacteria).…”

“…The genome scanning has been frequently used to evaluate the false discovery rate of function prediction tools [113,114]. To have a comprehensive understanding of methods' false discovery rate, the genomes of 4 model organisms representing 4 species kingdoms (homo sapiens from Animalia, arabidopsis thaliana from Plantae, saccharomyces cerevisiae from Fungi and mycobacterium tuberculosis from Bacteria)…”