PAN-CANCER ATLAS · 2026

A pan-cancer atlas of therapeutic T cell targets.

The most comprehensive catalogue of tumor-specific T cell targets to date — 16,687 empirically detected HLA-presented antigens across 21 cancer types, spanning 11 classes of molecular aberrations and identified in 88% of analyzed tumors.

16,687
Tumor-specific pHLAs
21
Cancer types
11
Aberration classes
88%
Tumors w/ targets
1,823
Immunopeptidomes
7,188
Tumor RNA-Seq
17,384
Normal controls
2,401
Neoantigens
Atlas at a glance

A pan-cancer view of therapeutic T cell targets

Hover any pin on the human-body map to preview that cancer, or click to jump straight to its filtered view in the explorer below.

Hover any pin to explore, or click to jump to that cancer in the atlas.
Target explorer

Search, filter, and rank the full atlas.

Every row is a tumor-specific peptide with its gene of origin, aberration class, HLA alleles, recurrence, abundance, tumor-vs-normal expression window, and DepMap essentiality. Click any row for the full annotation drawer.

All cancers

Showing 0 of 0 targets
Download CSV

Quick-start guide

  1. Select a cancer from the dropdown, or click a cancer tile/body-map pin above to filter by tumor type.
  2. Search by gene or peptide — type a gene name (e.g., PRAME, PMEL) or peptide sequence to find specific targets.
  3. Filter by class — click the colored chips (Canonical, Splicing, Mutation, etc.) to show only that aberration class.
  4. Sort any column — click column headers like Recurrence or Abundance to rank peptides. Click again to reverse.
  5. Click any row to open the detail drawer with MS/MS spectra, expression plots, HLA binding data, immunogenicity scores, and safety profiles.
  6. Hover anything — every column header, cancer code, class tag, and section title has an explanatory tooltip.
  7. Download — click "Download CSV" to export the currently filtered results with all annotations.

Keyboard: press Esc to close the detail drawer.

Jump to hero targets:
Cancer Peptide Gene Class Recurrence Abundance Spectral T/N window DepMap Homog. HLAs
Loading atlas…
All 21 cancers

21 cancers · 340,000+ ranked candidate targets

Each tile's donut (NC = non-canonical) shows the fraction of the repertoire from non-canonical sources — splicing, cryptic ORFs, retroelements, mutations, fusions, microbial. Click a tile to filter the explorer above.

Sorted by target count · click to filter explorer
Manuscript highlights

Hero targets uncovered by ImmunoVerse

Selected discoveries from the manuscript — each has experimental validation and, in several cases, direct evidence of T cell anti-tumor activity.

Alternative splicing · Melanoma

Splicing-derived PMEL (sPMEL)

KTWDQVPFSV · HLA-A*02:01

A novel PMEL peptide produced when exon 5 is skipped, detected in 89% of SKCM samples. Presented more abundantly than canonical PMEL targets with negligible expression in normal skin.

89%
SKCM recurrence
34%
HLA-A2 carriers
53%
Specific T cells
Transposable elements · Pan-cancer

LINE1 ORF2 as a pan-cancer target

RIAKSILSQK · ILPKVIYRF · SGYKINVQK

First evidence of native LINE1 ORF2 protein expression across 12 tumor types. Three peptides synthesized and confirmed by LC-MS/MS.

12
Tumor types
3
Validated peptides
Microbial antigens · Ovarian

N. circulans in 98% of ovarian tumors

Niallia circulans · C. ureolyticus

Recurrent tumor-resident microbes in gynecologic cancers. CMV peptides presented on HLA across 7 tumor types — first evidence of endogenous CMV epitopes in primary human tumors.

98%
OV tumors
48
CMV peptides
7
Tumor types
Cryptic ORFs · Non-canonical

Cryptic ORFs dominate non-canonical peptides

LYLETRSEF · AYPASLQTL

69% of the non-canonical immunopeptidome derives from cryptic ORFs. Ribo-Seq confirmed 62% ribosomal occupancy in neuroblastoma PDXs.

69%
Of non-canonical
62%
Ribo-Seq validated
Self-antigens · Pan-cancer

MAGE-A4 across 13 cancers; under-explored PRAME variant

GVYDGREHTV · GQHLHLETF

MAGE-A4 (FDA-approved TCR-T target) observed across 13 cancer types. PRAME GQHLHLETF detected in 67% of B*15:01 carriers vs 27% for the canonical SLLQHLIGL.

13
MAGE-A4 cancers
67%
PRAME recurrence (B*15)
Mutation neoantigens

2,401 empirically detected neoantigens

KRAS G12V · CTNNB1 T41A · NPM1 frameshift

2,401 immunopeptidome-evidenced neoantigens, including 127 public neoantigens shared across multiple tumor samples — of which 96% are not documented in the IEDB database. Covers well-known drivers (KRAS G12V, NPM1 frameshift, CTNNB1 T41A) plus newly identified shared neoantigens.

2,401
Total neoantigens
127
Public (shared)
96%
Of public, novel vs IEDB
Aberration classes

11 molecular sources of tumor-specific antigens

ImmunoVerse interrogates canonical self-antigens plus 10 non-canonical classes, each contributing distinct opportunities for cancer immunotherapy. Click any card to filter the explorer by that class.

Methods · Pipeline

How ImmunoVerse was built

An integrative multimodal pipeline combining transcriptomic events, tri-engine immunopeptidomic search, stringent multi-modal safety filtering, and single-cell essentiality analysis.

01
Catalogue
11 classes of tumor events

7,188 tumor RNA-Seq + 17,384 GTEx normals assembled into a histology-specific catalogue of canonical and non-canonical sources — filtered against 51 normal tissues.

02
Search
Tri-engine immunopeptidomics

1,823 immunopeptidomes searched with three complementary engines — MaxQuant, MS2Rescore, and the AI-driven Tesorai Search. Tesorai achieved up to 125% improvement in peptide IDs under high-confidence thresholds.

03
Safety
Multimodal safety screen

Candidates screened against the HLA Ligand Atlas (21 donors across 30 tissue types) and a Ribo-Seq healthy cryptic ORF database spanning 8 normal tissues — excluding an additional 19% (n=4,372) of candidates with normal-tissue evidence.

04
Binding
NetMHCpan 4.1 prediction

Predicted HLA binding layered onto empirical detection wherever HLA typing was available, excluding non-binders across all three search engines.

05
Prioritize
Therapeutic annotations

Each peptide scored on abundance, recurrence, normal-tissue exclusion, HLA affinity, DepMap essentiality, scRNA homogeneity, and MS confidence.

06
Deploy
Code-free cloud pipeline

Entire pipeline available as a code-free cloud workflow on Cancer Genomics Cloud — run your own data through ImmunoVerse in under an hour.

Data

Download the full atlas

Per-cancer JSON tables power this portal. Raw CSVs (with full abundance, HLA, and source annotations) are available through the main data repository — contact the corresponding author or clone the pipeline for the complete resource.

Per-cancer target tables

Each file contains every ranked candidate peptide for that cancer with the compact schema used by the explorer. For the complete annotated CSVs (including PSM-level intensities, NetMHCpan scores, and all matched HLA alleles), see the links below.

Full molecular catalogues ↗ HLA frequency & database ↗ Cancer Genomics Cloud pipeline ↗ GitHub repository ↗
Citation

Cite ImmunoVerse

If you use ImmunoVerse data, pipeline, or derived targets in your research, please cite the manuscript.

Open on bioRxiv ↗
Correspondence: Mark.Yarmarkovich@nyulangone.org · Yarmarkovich Lab · Perlmutter Cancer Center, NYU Grossman School of Medicine