Immunopeptides that are translated in cells and presented at the cell surface by major histocompatibility complex (MHC) mole-cules are important epitopes in basic Immunology and translational cancer immunotherapy. While most of the reported immuno-peptides are derived from protein coding sequence, the non-canonical peptides translated from “non-coding” regions are emerging and have attracted much attention in recent years. However, sensitive and accurate identification of such peptides remains a challenging task. Here we report an optimized approach integrating Ribo-seq and mass spectrometry to identify hundreds of non-canonical MHC-binding peptides. Three pipelines for analyzing Ribo-seq data were compared to generate small open reading frame (sORF) databases. Meanwhile, we have also combined bottom-up and de novo searching in proteomics data analysis and identified more immunopeptides. 7902 canonical and 308 non-canonical immunopeptides have been identified with selected ones vigorously validated. The present study provides a handy solution for identifying non-canonical MHC epitopes. The novel im-munopeptides resolving mechanisms of cancer antigen presentation, as well as applications in cancer immunotherapies.