The three Gencode ncORF tracks in the non-canonical ORF track container show non-canonical translated open reading frames (ncORFs) identified from ribosome profiling (Ribo-seq) data and mapped to the GENCODE annotation by the GENCODE / TransCODE consortium, as three tracks:
The Phase I catalog contains 7,264 unique human ncORFs called from Ribo-seq data across seven publications and mapped to GENCODE v35. Only translations of 16 codons or above and initiating from ATG start codons were incorporated. Redundant sense-overlapping ORFs were merged. Of these, 3,085 ORFs were found by more than one publication, providing independent replication evidence. This catalog was developed as part of an effort to standardize the annotation of translated ORFs across reference databases including Ensembl/GENCODE, HGNC, UniProtKB, and PeptideAtlas.
The Phase II catalog nearly quadruples the Phase I set, defining 28,359 ncORFs in the Comprehensive set, mapped to GENCODE v45. Compared to Phase I, additional published Ribo-seq datasets were incorporated and the 16-codon minimum length and ATG-only start codon restrictions from Phase I were removed.
Two subsets are provided for the Phase II data:
The default color scheme and available filter controls differ by track.
Kozak strength colors (default)
The Phase I track colors items by Kozak strength by default. Two alternative color schemes can be selected from the track controls page (Color by dropdown): Evidence type and HLA (Human Leukocyte Antigen) class (see below). Kozak strength is based on two specific positions in the genomic sequence, counted relative to the A of the ATG start codon: position −3 (three bases before the A) and position +4 (one base after the G).
| Golden amber: Strong match to the Kozak consensus. Both position −3 (A/G) and position +4 (G) match. | |
| Steel blue: Moderate match to the Kozak consensus. Only one of position −3 (A/G) or position +4 (G) matches. | |
| Gray: Weak match to the Kozak consensus. Neither position −3 (A/G) nor position +4 (G) matches. | |
| Black: Non-ATG start codon. The Kozak rule does not apply. | |
| Light gray: Genomic context could not be read (for example, an ORF too close to the edge of a chromosome). |
Evidence type colors (alternative)
Select Color by: Evidence type to color items by their peptide evidence from Deutsch et al. (see References). ORFs with no mass spectrometry evidence are gray.
| Gold: Annotated as a peptidein by Deutsch et al. (121 ORFs, Table S12). | |
| Steel blue: Detected only by HLA immunopeptidomics (1,667 ORFs). | |
| Forest green: Detected only by non-HLA (whole-cell tryptic) proteomics (51 ORFs). | |
| Orange: Detected by both methods (21 ORFs). | |
| Gray: No peptide evidence (5,404 ORFs). This includes ORFs whose only non-HLA peptides also map to known proteins and were excluded by the authors. |
HLA class colors (alternative)
Select Color by: HLA class to color items by whether their peptides were detected in HLA Class I, Class II, or both:
| Steel blue: Class I only (1,632 ORFs). | |
| Crimson: Class II only (10 ORFs). | |
| Orange: Both class I and class II (143 ORFs). | |
| Gray: No HLA data (5,479 ORFs). |
The Phase I track can be filtered by:
Mouseover for Phase I shows ORF name, host gene, Kozak strength and TE, replicated status, peptidein flag, HLA evidence tier, and HLA peptide count.
The Phase II Primary and Comprehensive tracks color items by Kozak strength using the same scheme as Phase I, but can only be filtered by start codon, Kozak strength, and Kozak TE, since peptide evidence fields are not included in the Phase II tracks.
Mouseover for Phase II shows ORF name, host gene, start codon, and Kozak strength and TE.
The raw data can be explored interactively with the Table Browser or the Data Integrator. The data can be accessed from scripts through our API; the track names are "gencNcOrfs" (Phase I), "gencNcOrfsPrimary" (Phase II Primary), and "gencNcOrfsComprehensive" (Phase II Comprehensive).
For automated download and analysis, the genome annotations are stored in bigBed files that can be downloaded from our download server. Individual regions or the whole genome annotation can be obtained using our tool bigBedToBed, which can be compiled from the source code or downloaded as a precompiled binary for your system. Instructions for downloading source code and binaries can be found here.
Mudge et al. (2022, see References) consolidated translation evidence from seven published ribosome profiling datasets that used harringtonine or lactimidomycin treatment to enrich for translation initiation sites. Ribo-seq reads were mapped to the GENCODE v35 annotation on GRCh38. Only ATG-initiated ORFs of at least 16 codons were retained, and redundant sense-overlapping ORFs were merged by taking the longest representative, yielding 7,264 ncORFs across seven biotype classes: upstream ORFs (uORFs), upstream overlapping ORFs (uoORFs), downstream ORFs (dORFs), downstream overlapping ORFs (doORFs), intronic ORFs (intORFs), lncRNA-embedded ORFs, and ORFs on processed transcripts. The catalog was developed as part of a reference-database coordination effort involving Ensembl/GENCODE, HGNC, UniProtKB, HUPO/HPP, and PeptideAtlas.
Chothani et al. (2025, see References) expanded the catalog by incorporating additional Ribo-seq datasets across more cell types and tissues and mapping to GENCODE v45. The ATG-start codon and 16-codon length restrictions were lifted to capture near-cognate initiations and micropeptides. A data-driven scoring framework using ribosome occupancy uniformity and P-site in-frame fraction identified a Primary subset of 10,127 ncORFs with translation signatures comparable to canonical coding genes; the Comprehensive set contains all 28,359 mapped ORFs.
Each ORF was annotated with its Kozak consensus strength by fetching the 11-base genomic context around the start codon from hg38.2bit and classifying positions −3 and +4 relative to the A of the start codon: both matching (A/G at −3 and G at +4) = Strong; one matching = Moderate; neither = Weak; non-ATG start = non-ATG. A numeric translational efficiency (TE) score was also assigned by looking up the 11-base context in the Noderer 2014 TE table (Mol Syst Biol 10:748, PMID 25170020).
Deutsch et al. (2026, see References) queried the 7,264 Phase I ORFs against two independent PeptideAtlas mass spectrometry repositories. The HLA immunopeptidomics build was constructed from HLA-I and HLA-II peptidomes across more than 100 HLA-typed donors spanning multiple tissue types and cancer cell lines; peptides were enriched by affinity purification and identified by tandem mass spectrometry. The whole-cell tryptic proteomics build used conventional shotgun proteomics from a broad range of cell lines and tissues. Spectra were manually reviewed and classified according to the Prensner et al. tier system (Tier 1B = numerous HPP-quality HLA peptides; Tier 2B = a single qualifying HLA peptide; Tier 1A/2A = additional non-HLA evidence). The study introduced the term peptidein for a translation product detectable by mass spectrometry but not yet annotatable as a protein due to absent functional evidence. The authors reviewed 137 of the 7,264 Phase I ORFs for annotation and classified 121 of them as peptideins (Table S12); a further 1,739 have HLA or tryptic peptide evidence but were not classified as peptideins.
The supplementary data tables (Tables S2, S3, S6, S7, and S12) from Deutsch et al. were
downloaded from the paper's supplementary materials at
https://doi.org/10.1038/s41586-026-10459-x.
Each table was joined to the Phase I bigGenePred by the short ORF identifier (e.g.,
c14riboseqorf80) using the script
addPeptideEvidence.py.
The script appended 14 new fields to all 7,264 Phase I items; the 5,353 ORFs that are not
in any of these tables receive default empty values so they remain visible in the track and
filterable on isPeptidein and related fields. Non-HLA peptides that map to known
proteins or are too short to be informative were excluded (Table S2, exclude
column), and the non-HLA peptide count includes only the remaining peptides.
The complete build procedure is documented in
ncOrfs.txt.
Thanks to Jonathan Mudge, Jorge Ruiz-Orera, John Prensner, Sebastiaan van Heesch, and the GENCODE / TransCODE consortium for creating and maintaining these annotations.
Deutsch EW, Kok LW, Mudge JM, Valls CF, Jungreis I, Ruiz-Orera J, Sun Z, Kusebauch U, Fierro-Monti I, Abelin JG et al. Expanding the human proteome with microproteins and peptideins. Nature. 2026 May 6;. PMID: 42092140
Chothani S, Ruiz-Orera J, Tierney JAS, Clauwaert J, Deutsch EW, Alba MM, Aspden JL, Baranov PV, Bazzini AA, Bruford EA et al. An expanded reference catalog of translated open reading frames for biomedical research. bioRxiv. 2025 Jul 7;. PMID: 40672165; PMC: PMC12265627
Mudge JM, Ruiz-Orera J, Prensner JR, Brunet MA, Calvet F, Jungreis I, Gonzalez JM, Magrane M, Martinez TF, Schulz JF et al. Standardized annotation of translated open reading frames. Nat Biotechnol. 2022 Jul;40(7):994-999. PMID: 35831657; PMC: PMC9757701