A reference translatome map reveals two modes of protein evolution

Abstract

AbstractRibosome profiling experiments demonstrate widespread translation of eukaryotic genomes outside of annotated protein-coding genes. However, it is unclear how much of this “noncanonical” translation contributes biologically relevant microproteins rather than insignificant translational noise. Here, we developed an integrative computational framework (iRibo) that leverages hundreds of ribosome profiling experiments to detect signatures of translation with high sensitivity and specificity. We deployed iRibo to construct a reference translatome in the model organism S. cerevisiae. We identified ∼19,000 noncanonical translated elements outside of the ∼5,400 canonical yeast protein-coding genes. Most (65%) of these non-canonical translated elements were located on transcripts annotated as non-coding, or entirely unannotated, while the remainder were located on the 5’ and 3’ ends of mRNA transcripts. Only 14 non-canonical translated elements were evolutionarily conserved. In stark contrast with canonical protein-coding genes, the great majority of the yeast noncanonical translatome appeared evolutionarily transient and showed no signatures of selection. Yet, we uncovered phenotypes for 53% of a representative subset of evolutionarily transient translated elements. The iRibo framework and reference translatome described here provide a foundation for further investigation of a largely unexplored, but biologically significant, evolutionarily transient translatome.