Updated project metadata. Ribosome profiling revealed translation outside of canonical coding sequences (CDSs) including translation of short upstream open-reading frames (ORFs), long non-coding RNAs, ORFs in UTRs or ORFs in alternative reading frames. Ribo-seq but also bioinformatics-based prediction and RNA-sequencing reported translation of thousands of ORFs derived from non-translated regions (NTRs). Although such ORFs gained increased attention over the years, their actual coding potential remains debated as protein products of only a fraction of them were identified by mass spectrometry. Here, we introduced a new workflow to discover translation products of NTRs at a large-scale. We combined reducing sample complexity (by enriching N-terminal peptides of cytosolic proteins as such peptides are ideal proxies for translation) with and extend search space (combining the sequences of UniProt proteins, UniProt isoforms and publicly available Ribo-seq data) reasoning that this combination increased chances of identifying proteins from NTRs. Further, we introduced rigorous data analysis and results curation workflows to cope with the increased complexity of the search space and to mine identified peptides. This stringent filtering approach was found essential to retain confident translational evidence at the peptide level for NTRs. We show that theoretically our strategy facilitates the detection of translation events of transcripts from NTRs, but experimentally less than 1% of all identified peptides might originate from such translation events.