uORF-seqr: A Machine Learning-Based Approach to the Identification of Upstream Open Reading Frames in Yeast

The identification of upstream open reading frames (uORFs) using ribosome profiling data is complicated by several factors such as the noise inherent to the procedure, the substantial increase in potential translation initiation sites (and false positives) when one includes non-canonical start codons, and the paucity of molecularly validated uORFs. Here we present uORF-seqr, a novel machine learning algorithm that uses ribosome profiling data, in conjunction with RNA-seq data, as well as transcript aware genome annotation files to identify statistically significant AUG and near-cognate codon uORFs.

A helicase links upstream ORFs and RNA structure

Upstream open reading frames (uORFs) in 5' UTRs of eukaryotic mRNAs are increasingly recognized as important elements that regulate cellular protein synthesis. Since uORFs can start from non-AUG codons, an enormous number of potential uORF initiation sites exists in 5'UTRs. However, only a subset of these sites is used and it has been unclear how actual start sites are selected. Studies of the DEAD-box helicase Ded1p from S. cerevisiae show that translation of uORFs with non-AUG initiation codons occurs upstream of mRNA structures that emerge with defective Ded1p. The data designate mRNA structure as important determinant for non-AUG initiation sites of uORFs. Ded1p can control this RNA structure and thereby regulate uORF translation.