There have been many reports showing translation of 5’ untranslated regions (5’-UTRs) and 3’ untranslated regions (3’-UTRs), mostly identified by ribosome profiling and/or tandem mass spectrometry (MS/MS) based proteomics. We propose a proteogenomic approach to identify UTR peptides from a MS/MS assay. Firstly, we construct a translated UTR peptide database with an assumption that UTR may be translated due to single nucleotide errors in recognizing START or STOP codon. After that, we apply a multi-stage search strategy (Madar et al., 2018), which is a method of rigorously identifying novel peptides. As a result, we identified 52 5’-UTR peptides and 9 3’-UTR peptides from a H1299 cell line dataset. There was a total of 45 and 9 genes corresponding the 5’-UTR and 3’-UTR peptides, respectively. Almost a half of 45 genes were commonly observed in a previous study. We further decided alternative start codon of 5’-UTR peptides based on codon frequencies, and then estimated the strength of its kozak context. We classified contexts into strong/weak/non-kozak classes (Lee et al., 2012). The kozak class composition of novel translation initiation sites (TISs) was comparable to that of the annotated translation initiation sites. As for read-through (RT) events at 3’ end of coding region, we identified a translation of 3’-UTR of MDH1 gene, which is consistent with a previous report (Stiebler et al., 2014). Furthermore, we could identify that the stop codon was substituted to tryptophan, which was not detected by Ribo-Seq. Finally, we also validated expression of 29 UTR peptides by MS/MS analysis of synthetic peptides. Among them, 28 UTR peptides were verified. Peptide identification using tUTR DB together with multi-stage strategy could rigorously identify UTR peptides translated due to single nucleotide errors.