Cancer-testis antigen design
Designing a cancer-testis antigen vaccine
Our aim is to design one vaccine antigen that targets cancer-testis antigens (CTAs) expressed across cancers responsible for the most deaths worldwide. We select CTA-specific protein regions with MS-observed MHC ligands, exclude shared non-CTA sequences and verified normal-tissue ligands, and assemble segments to broaden HLA coverage while limiting construct length and predicted junctional binding.
Explore the selected proteinsEach color follows one protein through the construct. Select a segment to explore its evidence.
What strict and loose actually allow
These are curated tissue-expression definitions. They do not establish that a tissue is immune privileged or that targeting it is safe.
Strict · core reproductive tissues
Testis, ovary and placenta. These tissues enter the core reproductive RNA fraction.
Loose · extended reproductive tract
The strict tissues plus cervix, endometrium, epididymis, fallopian tube, prostate, seminal vesicle and vagina.
Neither RNA numerator includes breast or lactating breast. OncoRef applies its curated reproductive-fraction and protein-expression rules, including exceptions; these are not absolute zero-expression gates. Thymus remains in the default RNA fraction denominator but is excluded from the somatic maximum and reproductive protein flag. Protein restriction conventions also use broader reproductive tissue labels. Each definition changes which protein isoforms count as non-CTA background.
CT83 · membership, evidence and selection
The proteins that made it through
Every target contributes at least one contiguous, CTA-specific piece containing an observed MHC ligand.
| Pieces | Peptide–HLA pairs | HLA alleles |
|---|
Identical full protein sequences share one target. Peptides shared between different targets count once in the overall total.
View the full protein sequence funnel
Which cancers drive the priority?
A protein scores higher when it is highly expressed in more samples from cancers that account for more deaths worldwide.
What “p95 occurrence” means
The percentage of cancer samples in which a protein ranks in the top 5% of expression within that sample.
For example, 30% means the protein reaches that expression rank in 30 out of 100 measured samples. It does not mean the vaccine covers 30% of patients.
world mortality share ×
p95 sample prevalence
The ranking adds these contributions across cancer categories. Incidence is shown for context and is not multiplied into the score.
Compare every protein across the cancer categories
Observed peptides, carefully assigned presenters
Mass spectrometry establishes that a peptide was observed. HLA restriction may be measured or inferred from binding predictions.
How the retained evidence is supported
Counts are distinct peptide–HLA pairs, assigned their strongest available evidence tier. Multiple alleles can share one observed peptide.
The inclusion rule
A longer observed peptide does not support an unobserved shorter sequence. HLA reach uses measured and inferred assignments as explicitly labeled; MS observations do not validate predicted HLA restrictions or establish T-cell recognition.
| HLA allele | Distinct peptides | Proteins | Monoallelic MS | Typed-sample inference | Untyped inference |
|---|
What each addition contributes
Compare allele-carrier reach, cancer-expression bounds and observed ligand evidence as proteins and segments are added.
Explore a cumulative protein panel
| Cancer | p95 expression union bounds | Unknown protein measurements |
|---|
HLA reach and cancer-expression coverage are separate calculations. Their curves are not multiplied into a clinical coverage claim. Expression cohorts are not a representative worldwide patient sample, and antigen processing, tumor HLA loss and immune responses are not included. The cancer axes use global burden shares without renormalizing the represented cancer subset.
Why choose regions rather than the whole protein?
Compare the retained native sequence with shared-protein regions and actual peptide observations from cancer and normal samples.
| Exact MS peptide / native position | Sample type | Tissue / disease / cells | HLA restriction evidence | Retained? | Study |
|---|
“Cancer-only in this corpus” means no noncancer observation was found in the supplied query, not that a peptide is cancer-specific. Absence of an MS hit is not evidence of absence. Normal-tissue hits in retained regions remain visible and require review. Peptides shared among genes cannot identify their source gene from sequence alone. MS supports observation, not T-cell recognition or vaccine safety.
A single, contiguous antigen
Protein segments stay contiguous. Terminal context and their order are adjusted to reduce predicted junctional binding.
View the complete annotated construct map
From cancer priorities to retained sequence
The same biological rules apply to both definitions.
Rank protein identities
Keep CTA-specific regions
Remove every residue covered by an 8-amino-acid sequence found in an independent non-CTA proteoform. Retain the remaining regions as contiguous stretches.
Require observed ligand evidence
Assemble within the length limits
How to interpret this design
This is a research antigen design. Expression ranking, normal-tissue restriction, observed peptides and binding predictions do not establish clinical efficacy or safety. Remaining junctional predictions need review and experimental validation.
Mortality and incidence values are shares of reference totals. Absolute counts are not available for every category. Cancer sample overlap and clinical patient coverage are not calculated.
Where the data come from
Carrier calculations follow the locus-aware assumptions described by Bui and colleagues. The worldwide donor-registry frequency reference is CIWD 3.0, as tabulated in Table A2 of Lőrincz and colleagues. These donors are a frequency proxy, not a census-weighted worldwide population.
Take the details with you
Download sequences, scientific figures and the complete tables for the selected definition.