Immunopeptidomics directly profiles peptides bound to human leukocyte antigens, or HLA molecules, capturing the antigen repertoire displayed on tumor-cell surfaces. The study’s authors identify several constraints in existing immunopeptidomics-based approaches to immunogenic peptide discovery: heavy reliance on sequence databases, dependence on matched normal controls and a lack of dedicated modules for predicting T-cell recognition. These limitations can restrict de novo peptide discovery, clinical applicability and the accuracy of immunogenicity prioritization, according to the research.

MSIPep is designed to address these gaps through an integrated analysis workflow. It combines database-based peptide searching with de novo sequencing strategies and supports both data-dependent acquisition, or DDA, and data-independent acquisition, or DIA, modes. The pipeline also applies false-discovery-rate control at multiple levels to improve the robustness of its results. Its source code is publicly available through a GitHub repository, and the developers provide a Docker container intended to support reproducible and scalable deployment.

For the prioritization stage, MSIPep incorporates ProTCR, an in-house deep-learning model developed to evaluate interactions between peptides and T-cell receptors. Rather than ranking candidates only by whether an HLA-bound peptide has been identified, the model provides a computational estimate of a peptide’s T-cell response potential. The combined framework therefore links identification of tumor-associated peptides with a receptor-interaction model intended to distinguish candidates with higher predicted immunogenicity, although that ranking remains a model-based estimate rather than a direct measurement of patient immune responses.

The evaluation used cancer datasets rather than a clinical trial. Acute myeloid leukemia applications highlighted MSIPep’s reported improvement in finding peptides with high predicted immunogenic potential, and analyses involving high-grade serous ovarian carcinoma and triple-negative breast cancer showed that the workflow could also be applied to solid-tumor data. These results position MSIPep as a reusable computational framework for narrowing large peptide-candidate spaces, but the supplied findings do not establish that prioritized peptides induce immune responses in people, validate them as cancer-vaccine targets or produce clinical benefit. The reported advance is therefore in the identification and computational prioritization of candidate immunogenic peptides across leukemia and solid-tumor datasets.

Subject of Research: Computational identification and prioritization of immunogenic tumor peptides

Article Title: MSIPep: A pipeline for immunogenic peptide identification and prioritization integrating immunopeptidomics and transcriptomics data

Article References: Liu, P., Zhang, S., Lu, M., Xu, M., Liu, Z., Chen, L., Xu, P., Tan, X., & Xie, L. (2026). MSIPep: A pipeline for immunogenic peptide identification and prioritization integrating immunopeptidomics and transcriptomics data. BMC Genomics. https://doi.org/10.1186/s12864-026-13481-5

Image Credits: AI Generated

DOI: 10.1186/s12864-026-13481-5

Keywords: MSIPep, immunopeptidomics, transcriptomics, cancer peptides, neoantigens, T-cell receptors, deep learning, mass spectrometry, acute myeloid leukemia, solid tumors, pipeline, immunogenic

The motivation behind the work was elegant in its simplicity. Somatic mutations in the CALR gene produce oncogenic calreticulin variants that drive certain myeloproliferative neoplasms by binding the thrombopoietin receptor MPL and activating it without ligand. These variants lose their KDEL endoplasmic reticulum retention signal, allowing them to be secreted or displayed on the cell surface. Because calreticulin was believed to be a lectin-like chaperone within the peptide loading complex of MHC class I, and because previous reports had documented decreased MHC class I surface expression, defective peptide loading, and an altered presented repertoire in cells expressing oncogenic calreticulin, the authors hypothesized that oncoCALR might be a uniquely sinister oncogene, one that simultaneously induces malignant transformation and immune evasion by corrupting the very machinery the immune system uses to detect it.

To test this hypothesis, the team engineered the murine pro-B cell line Ba/F3, derived from an H-2Kb-negative C3H mouse strain, to express murine MHC class I molecules fused to a mCherry reporter carrying ovalbumin-derived peptides of high, medium, and low binding affinity. These target cells, carrying either wild-type calreticulin, a calreticulin knockout, heterozygous or homozygous oncogenic mutations, or deletions of exon 9, were co-cultured with OT-I T cells that recognize the ovalbumin peptide in the context of H-2Kb. Flow cytometric assessment of the T cell activation marker CD69 after 6 and 24 hours revealed no defects at any time point in any mutant group. Even removal of the N-linked glycosylation sites N86 and N176 from H-2Kb, which should abolish the lectin-mediated interaction between calreticulin and the nascent MHC class I molecule, left antigen recognition by T cells entirely intact.

The researchers then extended their analysis to primary cells. Haematopoietic stem cells harvested from mice carrying oncogenic calreticulin or a calreticulin knockout in their haematopoietic system were lentivirally transduced to present the ovalbumin peptide by their endogenous H-2Kb complex and co-cultured with OT-I T cells for 72 hours. T cell activation, measured as the percentage of CD8-positive CD25-positive cells, and target cell killing were indistinguishable across all genotypes. The team did confirm previous reports of lower surface H-2Kb expression in homozygously oncoCALR-expressing and calreticulin-deficient cells, but this quantitative reduction did not translate into any functional impairment, consistent with the observation that even a single MHC class I molecule can trigger a T cell response through the MHC/TCR interaction.

The most ambitious test came in vivo. The researchers infected mice carrying heterozygous or homozygous oncogenic calreticulin, a calreticulin knockout, or wild-type calreticulin in their haematopoietic system, or in the liver under an AlbCre promoter, with 2 million focus-forming units of lymphocytic choriomeningitis virus strain Armstrong. Eight days post-infection, at the peak of the adaptive immune response, flow cytometric analysis of blood and spleen revealed no significant differences in the relative abundance of effector T cells or LCMV-specific effector T cells across genotypes. Cytokine production of splenic CD8-positive T cells restimulated with the LCMV gp33 peptide was unaltered in calreticulin knockout cells and, remarkably, significantly increased in heterozygous oncoCALR cells, while the cytotoxic protein granzyme B remained steady in all groups. The absence of calreticulin or the presence of its oncogenic variant did not compromise the antiviral T cell response in any organ system examined.

Finally, the team turned to human samples. CD34-positive haematopoietic stem cells were isolated from ten HLA-A02:01-positive and oncoCALR-positive myeloproliferative neoplasm patients and eight HLA-A02:01-positive healthy control donors. These target cells were co-cultured with a Jurkat T cell line engineered to express a chimeric antigen receptor recognizing the endogenous WT1-derived HLA-A*02:01-restricted peptide RMFPNAPYL. WT1 was selected as the antigen because it is equally expressed by healthy donor and patient-derived CD34-positive stem cells. CAR Jurkat activation, measured by CD69 expression after 48 hours, showed only a marginally lower response to patient-derived cells, and the researchers could find no differences in HLA surface expression, WT1 antigen levels, or expression of the T cell-suppressing ligand PD-L1 that would explain this slight decrease. MHC class I function in oncoCALR-expressing human cells was intact or, at most, minimally impaired.

Having failed to demonstrate a pivotal role for calreticulin in MHC class I functionality, the researchers sought to identify what calreticulin actually does at the proteome level. They performed tandem mass tag mass spectrometry on primary murine neutrophil granulocytes and cytotoxic T cells from calreticulin knockout, homozygous oncoCALR, and wild-type mice, restricting their analysis to proteins of the secretory pathway that could potentially interact with the endoplasmic reticulum-based chaperone. Five proteins were absent from samples of both test genotypes and both cell types: acyloxyacyl hydrolase, eosinophil peroxidase, neutral alpha-glucosidase AB, myeloperoxidase, and phospholipidase B-like 1. The abundance of murine MHC class I, H-2Kb, remained steady in the proteome despite its decreased surface expression, likely because the molecule accumulates in the Golgi apparatus rather than being degraded. These findings establish calreticulin as an indispensable chaperone for a specific subset of secretory pathway proteins, but not for MHC class I.

How, then, did calreticulin’s role in MHC class I function become so massively overestimated? The authors point to the history of its discovery. Calreticulin was found associated with MHC class I through co-immunoprecipitation assays from human cell lines in 1996 and later identified as part of the peptide loading complex by cryo-electron microscopy of affinity-purified extracts. The researchers argue that calreticulin, as a lectin chaperone capable of interacting with any glycoprotein bearing a Glc1Man9GlcNAc2 oligosaccharide, likely contaminates affinity purification-based analyses nonspecifically. Indeed, calreticulin is a member of the so-called CRAPome, a repository listing binding partners frequently found in co-immunoprecipitation and mass spectrometry protein interaction studies. Calreticulin may maintain a physical association with MHC class I, but this interaction does not influence its function. Moreover, some HLA alleles have been reported to be completely independent of the entire peptide loading complex.

The implications of this work are considerable. For patients with myeloproliferative neoplasms, the finding that MHC class I functionality is intact in oncoCALR-expressing cells encourages the feasibility of cancer vaccination approaches that previous studies had abandoned, claiming defects in antigen presentation as a cause. The identification of myeloperoxidase and other antibacterial immune proteins as calreticulin-dependent chaperone clients suggests that patients carrying oncogenic calreticulin mutations may harbor a subtle immune defect in the antibacterial response, an area that warrants further investigation. Calreticulin itself remains biologically essential: pan-calreticulin-deficient mice are not viable due to a defect in cardiogenesis, and female mice expressing oncogenic calreticulin homozygously in their haematopoietic system do not breed. The functional role of calreticulin in the peptide loading complex, the authors conclude, now needs to be redefined and its significance reevaluated, a conclusion that will force immunologists to reconsider a molecular relationship they had taken for granted since the mid-1990s.

Subject of Research: The role of calreticulin and its oncogenic variants in MHC class I antigen presentation and T cell immune responses

Article Title: Haematopoietic Cells Deficient for Calreticulin or Expressing Its Oncogenic Variants Maintain Functional Antigen Presentation

Article References: Höhrhan, M., Antonio‐Herrera, L., Colaço, H. G., Wagner, A., Unterwurzacher, M., Varga, C., Kratochwill, K., Bergthaler, A., & Kralovics, R. (2026). Haematopoietic Cells Deficient for Calreticulin or Expressing Its Oncogenic Variants Maintain Functional Antigen Presentation. Journal of Cellular and Molecular Medicine, 30(19), Article e71399. https://doi.org/10.1111/jcmm.71399

Image Credits: AI Generated

DOI: 10.1111/jcmm.71399

Keywords: calreticulin, MHC class I, antigen presentation, myeloproliferative neoplasms, oncogenic CALR mutations, T cell response, peptide loading complex, immunopeptidomics, cancer vaccination, chaperone biology, Haematopoietic, Cells

Much of modern cardioimmunology has been built on the premise that inflammation matters, but that its specificity does not. That premise has produced real clinical wins. Large trials of canakinumab, an interleukin-1beta antibody, demonstrated that blunting innate inflammation reduces cardiovascular events in patients with atherosclerotic disease, and low-dose colchicine has entered routine secondary prevention. More recently, randomized phase 2 trials of low-dose interleukin-2, intended to expand regulatory T cells, have tested whether boosting immune tolerance can calm vascular and myocardial inflammation after acute coronary syndromes. Imaging probes that track CCR2-positive monocytes into the injured heart have given clinicians a window onto the inflammatory infiltrate. Yet all of these approaches share a fundamental limitation: they target broad inflammatory programs rather than the antigen-specific lymphocytes that may be orchestrating, sustaining or sometimes protecting the myocardium.

The evidence that antigen-specific T cell responses to the heart exist is not new. In the late 1980s, researchers showed that immunizing genetically predisposed mice with cardiac myosin, the molecular motor of the heartbeat, induced autoimmune myocarditis, and subsequent work established that the disease was T cell mediated. The myocarditis-inducing epitope of cardiac myosin heavy chain alpha was mapped and shown to bind stably to major histocompatibility complex class II molecules on antigen-presenting cells. Decades later, the clinical relevance of these findings crystallized around a dramatic iatrogenic disease: myocarditis triggered by immune checkpoint inhibitors, the cancer therapies that unleash T cells against tumors. Independent teams showed that T cells specific for alpha-myosin and other cardiac myosin epitopes expand in patients with checkpoint-inhibitor myocarditis, providing the clearest demonstration yet that a heart antigen can drive fulminant human cardiac autoimmunity.

Why would the immune system be primed against its own heart? Part of the answer lies in central tolerance. The thymus deletes T cells that react strongly with self-proteins it displays, but the gene encoding alpha-myosin heavy chain, the dominant cardiac contractile protein, is expressed poorly or not at all in thymic medullary epithelial cells. This gap in thymic tolerance, demonstrated in both mice and humans, leaves a reservoir of T cells capable of recognizing the heart’s most abundant autoantigen. Peripheral mechanisms normally keep these cells in check, but injury can break the truce. When cardiomyocytes die, as in a heart attack or viral infection, cardiac proteins are released, taken up by dendritic cells and presented to T cells in a pro-inflammatory context, a process known as cross-priming. Molecular mimicry adds another route: microbial peptides that resemble cardiac epitopes, whether from group A streptococci in rheumatic heart disease or from the gut microbiota in experimental inflammatory cardiomyopathy, can activate T cells that then attack the heart.

The catalogue of validated cardiac antigens remains short but instructive. Cardiac myosin heavy chain alpha dominates, with defined class II-restricted epitopes and, more recently, identified class I epitopes for CD8 T cells. Cardiac troponin I is a second major player; unlike troponin T, it induces severe autoimmune inflammation in the myocardium when used to immunize mice, and autoantibodies against it correlate with outcomes after infarction. The beta-1 adrenergic receptor, long known as a target of functional autoantibodies in dilated cardiomyopathy, also contains epitopes that elicit T cell responses of varying pathogenicity. Additional candidates include the adenine nucleotide translocator, SERCA2a, the calcium pump of the sarcoplasmic reticulum, and branched-chain alpha-ketoacid dehydrogenase kinase. Beyond classical self-antigens, modified self can be antigenic: isolevuglandin-modified cardiac proteins drive CD4 T cell activation and dysfunction in the heart, and citrullination of myofilament proteins may reveal cryptic epitopes, a mechanism familiar from rheumatoid arthritis.

What these antigens share, and what the review proposes as a working framework, is a combination of abundance, cardiac enrichment and presentability. The heart is overwhelmingly composed of cardiomyocytes packed with contractile machinery, so its proteome is dominated by sarcomeric proteins that are scarce elsewhere. Proteomic atlases of the human heart, from quantitative maps of ventricular and atrial regions to single-cell and spatial transcriptomics of cardiac niches, now provide the raw material for predicting which proteins are sufficiently heart-enriched and sufficiently abundant to serve as targets. The critical filter is immunological: a candidate antigen must actually be processed and presented on MHC molecules. Strikingly, myosin-class II complexes have been detected in normal myocardium even before any injury, suggesting that the heart constitutively displays its own dominant antigen and that disease depends on context rather than mere exposure.

Not all heart-directed T cell responses are destructive, and this nuance may prove therapeutically decisive. Myocardial infarction has been shown to trigger cardioprotective, antigen-specific T helper responses, and a unique population of cardiac regulatory T cells potentiates protection after ischemic injury. In atherosclerosis, T cells reactive against apolipoprotein B begin as protective regulatory cells and can lose that identity as disease progresses, converting from guardians to drivers of plaque inflammation. Regulatory T cells that recognize troponin or myosin epitopes modulate healing after infarction, and the myocardial milieu itself favors local differentiation of regulatory T cells. The same antigen, in other words, can support repair or pathology depending on the functional polarization of the T cells that recognize it, a duality that any antigen-specific intervention must navigate carefully.

Turning this biology into medicine will require tools to see the relevant T cells, and such tools are finally emerging. MHC multimers loaded with defined cardiac peptides allow direct staining and isolation of antigen-specific T cells from blood and tissue. Single-cell T cell receptor sequencing, applied to paired atherosclerotic plaques and blood, has revealed expanded effector clones with autoimmune features, and comparable approaches have mapped the immune response in checkpoint-inhibitor myocarditis across heart, blood and tumor. Curated databases of pathology-associated TCR sequences, machine-learning methods for inferring TCR specificity, and CRISPR-based orthotopic TCR replacement for validating candidate receptors together form a toolkit that the review positions as the foundation for the next decade of mechanistic work. Transgenic mice carrying myosin-specific T cell receptors, first generated more than a decade ago, continue to illuminate how CD4 and CD8 subsets contribute to myocarditis and its progression to dilated cardiomyopathy.

The therapeutic horizon is equally provocative. Engineered T cells targeting fibroblast activation protein have been shown in preclinical studies to blunt cardiac fibrosis, including chimeric antigen receptor T cells produced in vivo via mRNA delivery, hinting at a future where the antigen-specific arm of immunity is redirected rather than suppressed. Conversely, costimulation blockade with abatacept has been deployed as rescue therapy for severe checkpoint-inhibitor myocarditis, a rational application of the two-signal model of T cell activation to a fulminant, antigen-driven disease. Preventive vaccine-style approaches that tolerize against cardiac autoantigens have shown promise in experimental heart failure. None of these strategies can be designed rationally, the review concludes, without a complete inventory of cardiac antigens and the T cell repertoires that recognize them. By fusing cardiac proteomics with immunopeptidomics, the systematic analysis of peptides displayed on MHC molecules, and with T cell receptor tracking from model organisms to patients, the field aims to convert heart-specific immunity from a scattered collection of observations into a defined, druggable axis of cardiovascular medicine.

Subject of Research: Antigen-specific T cell responses to cardiac proteins in myocarditis, myocardial infarction and heart failure

Article Title: Heart-specific T cell responses: from antigens to disease mechanisms

Article References: Campos Ramos, G. (2026). Heart-specific T cell responses: from antigens to disease mechanisms. Nature Reviews Cardiology. https://doi.org/10.1038/s41569-026-01356-x

Image Credits: AI Generated

DOI: 10.1038/s41569-026-01356-x

Keywords: T cells, cardiac antigens, myocarditis, myocardial infarction, heart failure, autoimmunity, MHC, immunopeptidomics, cardioimmunology, regulatory T cells, immune checkpoint inhibitors, T cell receptor

The technique at the heart of the study is called immunopeptidomics, and it represents a fundamentally different way of hunting for vaccine antigens. Traditional approaches have typically started with the genome or the proteome of a pathogen and then tested candidate proteins one by one to see whether T cells respond to them. Immunopeptidomics inverts that logic. Instead of asking what a pathogen might theoretically present, it asks what infected cells actually present. When a cell is infected by an intracellular bacterium such as F. tularensis, bacterial proteins are degraded inside the cell into short peptide fragments. A subset of those fragments is loaded onto major histocompatibility complex class II, or MHC-II, molecules and shuttled to the cell surface, where they can be surveyed by CD4 T cells. By isolating these MHC-II molecules from infected cells and sequencing the peptides bound to them with high-resolution mass spectrometry, researchers obtain a direct molecular snapshot of the antigen display process itself.

In the new work, the team infected dendritic cells, the professional antigen-presenting cells of the immune system, with the live vaccine strain of F. tularensis, known as LVS. This attenuated strain has been used for decades and provides partial protection against fully virulent strains in animal models, making it a useful system for dissecting the immune response it induces. After allowing time for antigen processing and presentation, the researchers purified MHC-II peptide complexes from the infected cells and analyzed them by liquid chromatography-tandem mass spectrometry. The screen yielded 57 distinct MHC-II peptides derived from 23 different F. tularensis proteins. This is a technically demanding achievement. Bacterial peptides are vastly outnumbered on the cell surface by peptides derived from the host cell’s own proteins, so detecting dozens of pathogen-derived sequences requires careful enrichment, sensitive instrumentation, and rigorous validation to distinguish genuine bacterial ligands from background contamination.

One of the most striking findings to emerge from the peptide catalog was its composition. When the authors cross-referenced the 23 source proteins against existing proteomic data for F. tularensis, they found that most of them belong to the highly expressed fraction of the bacterial proteome. This observation carries real biological weight. Antigen presentation is not a random sampling of everything a pathogen makes; it is shaped by how much of each protein the bacterium produces, how quickly the protein is degraded inside the host cell, and how efficiently the resulting fragments fit into MHC-II binding grooves. The predominance of abundant proteins among the presented peptides suggests that the antigen display machinery preferentially samples the proteins that the bacterium relies on most heavily during infection, which is precisely the subset of molecules a vaccine would ideally target.

Identifying presented peptides, however, is only the first step. A peptide displayed on an infected cell is a candidate antigen, but whether the immune system actually mounts a measurable T cell response against it is a separate question. To address this, the researchers immunized mice with the LVS strain and then tested whether CD4 T cells from the immunized animals responded to each of the identified peptides. The readout they used was the production of interferon-gamma, a signature inflammatory cytokine released by activated CD4 T cells of the Th1 type, which is the helper subset considered most relevant for controlling intracellular bacteria. Out of the full panel of peptides, ten emerged as bona fide immunogenic epitopes, meaning they reliably triggered interferon-gamma responses in T cells from vaccinated mice. The breadth of this response was itself informative, revealing that the immune system’s attention after tularemia vaccination is spread across multiple bacterial proteins rather than concentrated on a single dominant target.

Within that set of ten epitopes, a clear hierarchy became apparent. Five peptides, derived from the antigens FTL_1678, DacD, DnaK, FTL_1498, and MinD, stood out as dominant epitopes, each eliciting a response in at least 80 percent of the immunized animals. The identities of these proteins are intriguing. DnaK is a well-characterized molecular chaperone, a highly conserved protein that helps other bacterial proteins fold correctly, and its abundance inside the bacterium likely explains both its prominence in the MHC-II peptide pool and its immunogenicity. DacD is an enzyme involved in bacterial cell wall metabolism, MinD participates in the machinery that positions the cell division plane, and FTL_1678 and FTL_1498 are proteins of less certain function annotated by their gene numbers in the F. tularensis genome. The fact that two of the five dominant targets are proteins of unknown function underscores how much of the protective immune response to this pathogen has remained invisible to researchers until now.

The logical next question was whether these newly discovered epitopes could be assembled into something therapeutically useful. The team constructed a peptide vaccine containing the dominant epitopes and tested it in mice. The results were encouraging but nuanced. The peptide formulation proved immunogenic, successfully priming CD4 T cell responses against the included epitopes. When vaccinated mice were subsequently challenged with the fully virulent F. tularensis FSC200 strain, the vaccine contributed to the early control of bacterial replication, reducing the bacterial burden during the initial phase of infection. However, the epitope cocktail alone was not sufficient to confer complete protection against the virulent challenge. This outcome is consistent with a growing recognition in the tularemia field that effective immunity against this organism likely requires the coordinated action of multiple immune mechanisms, including CD4 T cells, CD8 T cells, antibodies, and innate effector programs, and that a peptide vaccine targeting a handful of CD4 epitopes represents one component of a more complex protective recipe rather than a standalone solution.

Even with that caveat, the practical implications of the study extend well beyond the immediate vaccine result. The identified epitopes provide researchers with precise molecular reagents for the first time. With known peptide sequences and the knowledge that they are both naturally presented and immunogenic, investigators can now build tetramer or other staining tools to directly count and characterize antigen-specific CD4 T cells in blood and tissues during vaccination studies. Such tools have transformed the study of T cell biology in viral systems, where epitopes are well mapped, but have been largely unavailable for F. tularensis. The epitopes could also serve as benchmarks for comparing vaccine candidates, allowing researchers to measure whether a new formulation actually expands responses to the same protective targets that the live vaccine strain induces. In this sense, the study delivers a toolkit as much as a vaccine lead.

Methodologically, the work also offers a template that can be applied to other intracellular pathogens where the antigenic targets of protective T cells remain obscure. Immunopeptidomics sidesteps the biases of prediction algorithms, which rely on inferred binding motifs and often miss real ligands, and instead reports the actual peptide repertoire as processed by living cells. As mass spectrometry sensitivity continues to improve, the approach is likely to reveal progressively deeper layers of the pathogen-derived peptidome, including peptides from low-abundance proteins and noncanonical sources. For F. tularensis specifically, the identification of conserved and abundant proteins as dominant CD4 T cell targets raises the possibility that immune recognition focuses on functions the bacterium cannot easily alter without fitness costs, a property that would make these antigens comparatively stable vaccine components.

The study leaves open questions that will shape the next phase of research. Whether the same epitopes are presented and recognized in the context of infection with fully virulent strains, rather than the attenuated LVS used for screening, remains to be confirmed, as does their relevance to human immune responses, since the current work was conducted in mice with mouse MHC-II molecules. It also remains to be determined which combinations of epitopes, delivery formats, and adjuvants can convert the early control of bacterial replication observed here into durable, complete protection. What the study establishes, however, is a concrete starting point: a validated list of the bacterial proteins that infected cells actually display, and a demonstration that several of them drive reproducible CD4 T cell immunity. For a pathogen that has resisted vaccine development for generations, that list may prove to be the most valuable map produced in the field in years.

Subject of Research: Identification of immunogenic CD4 T cell epitopes of Francisella tularensis by MHC-II immunopeptidomics in mice

Article Title: Immunopeptidomics reveals conserved and abundant Francisella tularensis proteins as immunogenic CD4 T cell antigens in mice

Article References: Laskova, P., Porkertova, S., Ballek, O., Balonova, L., Pavloskova, J., Neuwirth, A., Pavkova, I., Fabrik, I., Lukac, P., Vannucci, L., Filipp, D., Stulik, J., & Link, M. (2026). Immunopeptidomics reveals conserved and abundant Francisella tularensis proteins as immunogenic CD4 T cell antigens in mice. PLOS Pathogens, 22(9), e1014597. https://doi.org/10.1371/journal.ppat.1014597

Image Credits: AI Generated

DOI: 10.1371/journal.ppat.1014597

Keywords: Francisella tularensis, tularemia, immunopeptidomics, CD4 T cells, MHC class II, epitopes, vaccine development, mass spectrometry, dendritic cells, interferon-gamma, biodefense, intracellular bacteria

The term dark proteome covers an astonishing diversity of molecular species. It includes proteins translated from non-canonical open reading frames — stretches of RNA that were long dismissed as non-coding filler — as well as products of repeat elements such as endogenous retroviruses and LINE-1 transposons, peptides derived from circular RNAs, mistranslated proteins produced when the ribosome inserts the wrong amino acid, proteasomally spliced peptides, and stress-induced translation products that appear only when cells are pushed to the brink. Over the past decade, an unprecedented depth of resolution has been attained in cataloguing this hidden world, with individual classes of dark proteins now numbering in the thousands of members each. A massive proteogenomic screen published in 2024, for example, identified thousands of novel peptides from the human dark proteome, and a 2026 study in Nature expanded the human proteome with microproteins and peptideins — molecules with evidence for their existence but ambiguous or absent evidence for physiological function in normal cells.

The technological engine behind this discovery wave is ribosome profiling, or Ribo-seq, a technique first developed in Weissman’s laboratory that captures snapshots of every ribosome engaged on every messenger RNA in a cell. Ribo-seq revealed that translation is far more pervasive than textbook models suggested: ribosomes initiate not only at annotated start codons but at hundreds of thousands of upstream and alternative sites, in long non-coding RNAs, in 5′ untranslated regions and even in pseudogenes. Complementing Ribo-seq, immunopeptidomics — the mass spectrometric analysis of peptides presented on the cell surface by major histocompatibility complex molecules — has shown that a striking fraction of the peptides displayed by cancer cells comes from these unannotated sources. Studies in melanoma, lung cancer and pancreatic cancer have demonstrated that noncoding and non-canonical regions are among the main sources of targetable tumour-specific antigens, fundamentally challenging the assumption that neoantigens arise mainly from point mutations in canonical genes.

What makes the dark proteome more than a curiosity is the accumulating evidence that its members do real work in cancer biology. Individual dark proteins have now been characterized as oncogenes, tumour suppressors, signalling regulators and cancer-specific genetic dependencies. The microprotein TPM3P9, encoded by a cryptic long non-coding RNA, drives oncogenic RNA splicing and tumorigenesis. The TINCR ubiquitin-like microprotein acts as a tumour suppressor in squamous cell carcinoma, while a related microprotein called pTINCR promotes epithelial differentiation and suppresses tumour growth through activation of the CDC42 protein. In glioblastoma, a peptide encoded by the upstream open reading frame of the MYC messenger RNA binds to the tropomyosin receptor kinase B and promotes tumour growth in mice, and the upstream-ORF-encoded MP31 disrupts mitochondrial quality control and inhibits tumorigenesis. In childhood medulloblastoma, translation of non-canonical open reading frames appears to function as a cancer cell survival mechanism in its own right.

Perhaps the most dramatic examples come from dual-coding genes — loci where a single stretch of DNA yields both a canonical protein and a second, unconventional product. The NCYM gene, a cis-antisense gene of MYCN, encodes a de novo evolved protein that inhibits GSK3β and thereby stabilizes MYCN in human neuroblastomas, a finding that has been linked to clinical outcome. The classic CDKN2A tumour suppressor locus encodes not only p16INK4A but also p14ARF, an alternative reading frame product that forms meso-scale assemblies upon phase separation with the nucleolar protein NPM1. Even the circular RNA world has yielded a functional oncogene: Firefox, a protein encoded by circular PVT1 RNA, was shown in 2026 to be essential for MYC-driven oncogenesis. These cases demonstrate that the dark proteome is not merely translational noise but a layer of biology with direct consequences for tumour evolution and treatment response.

How does cancer gain access to this hidden protein repertoire? The Perspective outlines several mechanisms of dysregulation. Oncogenic transcription factors, particularly chimeric fusion proteins characteristic of certain leukaemias and sarcomas, can drive tumour-specific transcription, processing and translation of silent genomic regions, effectively switching on genes that were never meant to be expressed. Extrachromosomal DNA — circular DNA molecules that carry amplified oncogenes such as MYC in pancreatic cancer — can generate fusion transcripts and aberrant translation products at high levels. Nutrient stress provides another route: when tumours experience tryptophan or arginine deprivation, the ribosome misincorporates substitute amino acids, generating so-called substitutants that are foreign to the immune system. Chemotherapy and leucine deprivation likewise induce codon-biased aberrant protein production. Meanwhile, oncogene-driven sloppiness in mRNA translation, ribosomal frameshifting and stress-response pathways such as the integrated stress response and the unfolded protein response all conspire to flood the cancer cell with non-canonical peptides, many of which end up displayed on the cell surface.

It is on the immunotherapy front that the dark proteome may deliver its most immediate clinical payoff. Because many dark proteins are encoded by genomic regions that are not expressed in healthy tissues, the immune system has never been trained to tolerate them, making their peptides potentially ideal tumour antigens. Endogenous retrovirus-derived antigens have already been shown to promote lung cancer immunotherapy, and antibodies against these ancient viral proteins correlate with response to checkpoint blockade. Shared epitopes from endogenous retroviruses can induce high-avidity cytotoxic T cells, and tumour circular RNAs have been shown to elicit anti-tumour immunity by encoding cryptic peptides. In pancreatic cancer, tumour-restricted cryptic antigens have been validated as targets for T cell recognition. The therapeutic modalities now in play span the full modern immunotherapy arsenal: personalized and shared vaccines, TCR-engineered T cells, CAR-T cell approaches and bispecific T cell engagers. A first-in-class bispecific T cell engager targeting a DARKFOX peptide presented by solid tumours has already entered preclinical development, and a clinical trial of a dark-proteome-directed immunotherapy is registered on ClinicalTrials.gov.

Turning dark proteins themselves into drug targets is harder, but not impossible. Many are small, intrinsically disordered molecules that lack the stable pockets conventional small-molecule drugs require. Yet structure prediction tools such as AlphaFold, together with new protein design methods for binding intrinsically disordered regions, are beginning to make the undruggable tractable. The MYC inhibitor omomyc — a miniprotein that exploits intrinsic cell-penetrating activity — has advanced to a phase 1 clinical trial in solid tumours, and a synthetic cell-penetrant miniprotein called DuoMYC efficiently inhibits oncogenic MYC. EMBOW-derived peptides targeting the WDR5-MLL1 complex have suppressed leukaemia progression in preclinical models. Functional genomics is doing the other half of the work: CRISPR-based screens have revealed that human cancer cells depend on cryptic non-canonical open reading frames, and large-scale short open reading frame screens have identified microproteins involved in cancer cell fitness, feeding into resources such as the cancer dependency map and the dark kinase knowledgebase.

The authors are careful to emphasize how much remains unknown. Key gaps persist in understanding how the dark proteome is regulated, how much it actually contributes to cancer phenotypes versus representing tolerated noise, how antigenic its products truly are in patients, and whether dark proteins can be developed into clinical-grade therapies with acceptable safety profiles. Community efforts to standardize the annotation of translated open reading frames, benchmark mass spectrometric detection of unannotated microproteins and expand reference catalogues such as GENCODE are underway, supported by initiatives including the Cancer Grand Challenges dark proteome teams. Spatial transcriptomics and single-cell Ribo-seq are adding tissue-level context, revealing where and when unconventional translation occurs in intact tumours. What is clear is that the dark proteome offers the possibility to advance cancer science, diagnosis and treatment in ways the genomics revolution alone could not. The darkness, the authors suggest, is not a void but an unexplored continent — and the maps are finally being drawn.

Subject of Research: The functional dark proteome in cancer, comprising non-canonical proteins and peptides with roles in tumour biology and immunotherapy

Article Title: Mapping the functional dark proteome in cancer

Article References: Prensner, J. R., Freed-Pastor, W. A., Abelin, J. G., Wu, C. J., Merbl, Y., Pilpel, Y., Schumacher, T. N., Weissman, J. S., van Heesch, S., & Agami, R. (2026). Mapping the functional dark proteome in cancer. Nature Reviews Cancer. https://doi.org/10.1038/s41568-026-00985-1

Image Credits: AI Generated

DOI: 10.1038/s41568-026-00985-1

Keywords: dark proteome, cancer, non-canonical open reading frames, microproteins, Ribo-seq, immunopeptidomics, neoantigens, cancer immunotherapy, CRISPR screens, tumour suppressors, oncogenes, translational dysregulation

The researchers focused on the 4T1 model, a highly aggressive mouse breast cancer cell line that is widely used to study triple-negative breast cancer, the subtype of the disease that lacks the three most common therapeutic targets and that has historically been difficult to treat with immunotherapy. Although checkpoint blockade has improved outcomes for some patients with triple-negative breast cancer when integrated into neoadjuvant treatment, the 4T1 model has been an underused resource for dissecting resistance mechanisms because, until now, no tumor-specific epitopes had been defined within it. Without knowing what peptides the immune system can recognize, experiments designed to test immunotherapy combinations in this model have lacked a molecular anchor.

To fill that gap, the team turned to immunopeptidomics, the mass spectrometry-based analysis of the repertoire of peptides presented on the cell surface by major histocompatibility complex molecules. This technique allows researchers to catalog, in an unbiased way, every fragment of protein that a cell displays to patrolling T cells. When the immunopeptidomes of different 4T1 subclones were profiled, a striking pattern emerged. The parental, less metastatic cells presented a broad array of peptides, whereas subclones selected for enhanced metastatic behavior showed a marked contraction of their presented peptide repertoire, consistent with a downregulation of the antigen processing and presentation machinery.

This loss of antigen presentation is a classic strategy of immune evasion, and the study went further by showing that the defect is not irreversible. When the metastatic subclones were treated with type I interferon, a signaling molecule central to antiviral and antitumor immunity, the machinery of antigen processing and presentation was restored, and the cells once again displayed a richer set of peptides on their surface. The finding suggests that metastatic cells can silence their own visibility to the immune system through a pathway that interferon signaling can counteract, a mechanistic insight with direct implications for how interferon-based approaches might be combined with checkpoint inhibitors.

The most consequential discovery, however, lay in the identity of the peptides themselves. Rather than being dominated by fragments of conventional genes mutated or overexpressed in cancer, the immune responses mounted against metastatic 4T1 cells were directed largely at cryptic peptides derived from endogenous retroviral products, particularly those encoded by murine leukemia virus, or MLV. These are sequences that most tumor immunology pipelines would discard as non-coding or repetitive noise, yet in this model they constitute the immunodominant landscape, meaning the targets that the T cell response preferentially recognizes and attacks.

Interferon treatment did more than simply restore presentation; it reshaped the retroviral target landscape itself. Exposure to type I interferon increased both the abundance and the diversity of MLV-derived peptides in the 4T1 immunopeptidome, effectively broadening the range of viral remnant fragments that tumor cells displayed. This interferon-driven amplification of retroviral antigen display provides a mechanistic bridge between innate antiviral signaling and adaptive tumor recognition, and it helps explain why interferon signaling has often been associated with better responses to immunotherapy in both mouse models and human cancers.

Crucially, the functional experiments demonstrated that these retroviral targets are not merely decorative. MLV-specific T cell responses proved to be critical for the effectiveness of immune checkpoint blockade in the model. When the researchers examined tumor growth and metastasis-free survival, the data showed that T cells recognizing the endogenous retroviral peptides were essential for suppressing tumor outgrowth and prolonging the period during which animals remained free of metastatic disease. In other words, the anti-metastatic immunity that determines outcome in this aggressive breast cancer model is, to a substantial degree, an immune response against the tumor’s own ancient viral passengers.

The implications extend beyond the mouse. Human genomes carry their own burden of endogenous retroviral elements, and human tumors frequently express retroviral proteins that are absent from healthy adult tissues. If the same logic applies in patients, then the retroviral peptidome of human tumors could represent a shared, immunodominant and potentially druggable antigen class, one that does not depend on the private mutations that make personalized neoantigen vaccines so logistically complex. The identification of defined, immunodominant ERV epitopes in the 4T1 model now gives researchers a concrete platform on which to build ERV-targeted immunotherapies, including peptide vaccines and T cell-based approaches, and to test them in a system where the relevant antigens are finally known.

The study also reframes the role of interferon in cancer immunotherapy. Rather than acting only as a general inflammatory amplifier, type I interferon emerges here as a specific regulator of the retroviral antigen repertoire, expanding the visible surface of the tumor to the immune system. The authors suggest that this provides a foundation for translating interferon and ERV-targeting strategies into future clinical trial design, potentially pairing agents that derepress endogenous retroviral expression or enhance interferon signaling with checkpoint blockade, so that the immune system is presented with a richer set of retroviral targets precisely when therapeutic antibodies release the brakes on T cells.

For a field that has struggled to explain why some triple-negative breast cancers respond to immunotherapy while others do not, the message of this work is that the answer may lie partly in the viral archaeology of the genome. Metastatic cells that hide their antigens escape; cells forced by interferon to display their endogenous retroviral fragments become visible and vulnerable. By naming the specific MLV-derived epitopes that dominate the anti-metastatic response in 4T1 tumors, the researchers have converted a widely used but immunologically opaque model into a defined testing ground for retrovirus-directed cancer vaccines, and they have strengthened the case that the ancient viral DNA within us is not silent baggage but an active participant in the fight against cancer.

Subject of Research: Identification of immunodominant endogenous retroviral peptides that drive anti-metastatic T cell immunity and checkpoint blockade response in a triple-negative breast cancer mouse model.

Article Title: Identification of immunodominant ERV peptides critical for anti-metastatic immunity in a model of TNBC

Article References: Faridi, P., Zanker, D., Haynes, N. M., Nickson, J., Spurling, A., Panetta, F., Huang, P., Sung, N., Kang, J.-H., Dolcetti, R., Rajapaksha, H., Robinson, B. W. S., Dick, I., Redwood, A., Creaney, J., Chen, W., Anderson, R. L., Caminschi, I., Purcell, A. W., & Parker, B. S. (2026). Identification of immunodominant ERV peptides critical for anti-metastatic immunity in a model of TNBC. Cancer Immunology, Immunotherapy. https://doi.org/10.1007/s00262-026-04554-1

Image Credits: AI Generated

DOI: 10.1007/s00262-026-04554-1

Keywords: endogenous retrovirus, ERV peptides, triple-negative breast cancer, 4T1 model, immunopeptidomics, murine leukemia virus, type I interferon, immune checkpoint blockade, antigen presentation, metastasis, cancer immunotherapy, tumor immunology

The biological logic behind neoantigen vaccines is compelling. Neoantigens arise from somatic mutations that exist only in tumor cells, which means the immune system has not been trained to tolerate them, and the risk of off-target autoimmunity is theoretically low. Advances in next-generation sequencing, HLA typing, immunopeptidomics and synthetic vaccine platforms have made it feasible to sequence a patient’s tumor, identify candidate mutation-derived antigens and manufacture an individualized vaccine in a matter of weeks. Early clinical studies have now established feasibility, safety and immunogenicity across a striking range of solid tumors, including melanoma, pancreatic ductal adenocarcinoma, hepatocellular carcinoma, non-small cell lung cancer, bladder cancer and renal cell carcinoma, using platforms that span mRNA, peptide, DNA, viral vector and dendritic cell technologies.

The clinical momentum is real. In the phase 2b KEYNOTE-942 trial, the individualized mRNA vaccine mRNA-4157, known as V940, combined with pembrolizumab achieved 18-month recurrence-free survival of 79 percent versus 62 percent with pembrolizumab alone in resected melanoma, a hazard ratio of 0.561, with no grade 4 or 5 adverse events attributed to the vaccine. In pancreatic cancer, one of the hardest immunologic targets in oncology, responders to the personalized RNA vaccine autogene cevumeran showed markedly prolonged recurrence-free survival compared with non-responders, with a median not yet reached versus 13.4 months. A phase 3 trial in resected high-risk melanoma, INTerpath-001, is now enrolling roughly 1,089 patients, a sign that the field is moving from proof of concept toward pivotal testing.

Yet beneath these headline numbers lies an uncomfortable attrition problem that the review dissects stage by stage. Only a small fraction of the mutations identified in a tumor ever become genuinely immunogenic vaccine targets. In the autogene cevumeran pancreatic cancer trial, just 11 percent of the administered vaccine neoantigens, 25 out of 230, elicited measurable T-cell responses. In the NeoVax melanoma study, roughly 60 percent of vaccine neoantigens induced CD4-positive T-cell responses but only about 16 percent triggered CD8-positive cytotoxic T cells. A recent systematic evaluation found that only around 22 percent of predicted neoepitopes are actually presented on HLA molecules. Each false positive dilutes the effective immunogenic content of a vaccine that has limited room for passengers.

The root of the problem is that antigen presentation is a multi-step biological process, involving proteasomal cleavage, TAP-mediated transport and peptide-HLA loading, that most computational pipelines compress into a single binding-affinity prediction. Even when a peptide is presented, whether it provokes a functional T-cell response depends on the three-dimensional docking geometry between the peptide-HLA complex and a cognate T-cell receptor, governed by molecular flexibility and peptide conformation that sequence-based models capture poorly. Compounding matters, the scarcity of paired TCR-peptide-MHC datasets and the staggering diversity of the TCR repertoire make generalization across patients extraordinarily difficult. Manufacturing timelines, typically seven to eight weeks for peptide and viral vector platforms, add a further constraint, since tumors can evolve while the vaccine is being built.

This is where the review sees artificial intelligence earning its place, not as a replacement for experiments but as a way to narrow enormous biological and chemical search spaces. At the discovery stage, machine learning frameworks now integrate genomic, transcriptomic and mass spectrometry-based proteomic data to prioritize candidates with multi-layer evidence of expression and presentation. Tools such as ProGeo-neo combine proteomics with patient-specific genomic profiles to confirm protein-level expression, while ImmuneMirror merges whole-exome and RNA sequencing data, and modular workflows like TIminer and pVACtools automate reproducible end-to-end immunogenomic profiling. The authors also highlight the growing importance of noncanonical antigens, derived from alternative open reading frames, aberrant splicing, intron retention, fusion transcripts, circular RNA products and endogenous retroviral elements, some of which may be shared across patients and could inform semi-shared vaccine strategies.

Presentation prediction is where measurable gains have been most convincing. Pan-allelic tools such as NetMHCpan integrate eluted ligand data with binding affinities, and MHCflurry 2.0 folds antigen-processing features into neural network models, achieving an area under the curve of roughly 0.911 for binding classification in a benchmark of 18 predictors across 32 HLA alleles. Newer architectures such as CapHLA jointly model class I and class II presentation, and protein language model-based approaches like MUNIS improve prioritization of presented epitopes. The review also makes a pointed case for CD4-positive T cells, which clinical studies have repeatedly activated, sometimes more frequently than CD8 responses, arguing that HLA class II prediction, despite its computational challenges from open binding grooves and variable peptide lengths, deserves far more attention in vaccine design.

At the T-cell recognition frontier, deep learning is pushing into structure. AlphaFold-based modeling of peptide-HLA and TCR-peptide-HLA complexes can discriminate true target epitopes from decoys, while tools such as NetTCR-2.0, DeepTCR, TRAP and SageTCR integrate paired TCR sequences and three-dimensional structural features to predict receptor binding. Benchmarking studies, however, consistently show that these models remain strongly data-dependent and often fail to generalize to unseen epitopes, so the authors position them as prioritization and hypothesis-generating tools rather than definitive immunogenicity predictors. Downstream, AI is also reshaping formulation science: the AGILE platform uses deep learning-guided screening to identify ionizable lipids for mRNA delivery, and the LiON framework applies message-passing neural networks to large lipid nanoparticle datasets, narrowing chemical design spaces before any wet-lab testing begins.

To tie these threads together, the authors propose an evidence-weighted, closed-loop decision-gate framework spanning six stages: candidate generation, tumor relevance filtering, HLA presentation evidence, immunogenicity refinement, portfolio-level selection, and manufacturability and clinical delivery, with post-vaccination immune monitoring data fed back to refine future models. They stress that repertoire size should reflect biologically meaningful coverage rather than numerical abundance, since no clinically validated optimal neoantigen count exists. On the regulatory front, the UK’s Medicines and Healthcare products Regulatory Agency has released draft guidance for individualized mRNA cancer immunotherapies, in which AI selection algorithms may qualify as Software as a Medical Device, requiring Good Machine Learning Practice, transparent pipelines, version tracking and barcode-based traceability linking genomic data to the administered dose. The review’s bottom line is measured but optimistic: AI will not conjure effective cancer vaccines on its own, but embedded within rigorous experimental validation, scalable manufacturing and evolving regulatory oversight, it may finally make personalized neoantigen vaccination a reproducible pillar of precision oncology.

Subject of Research: The application of artificial intelligence across the translational development pipeline of personalized neoantigen cancer vaccines, including neoantigen discovery, HLA presentation prediction, T-cell recognition modeling and formulation optimization.

Article Title: Artificial intelligence for translational personalized neoantigen cancer vaccine development

Article References: Wei, C.-Y., Lin, H.-C., Wu, C.-J., Kuo, C.-N., Chang, C.-M., Chou, W.-H., Chou, S.-P., Feng, S.-H., Huang, W.-C., Jiang, S., Fairfax, B. P., Lee, K.-Y., & Chang, W.-C. (2026). Artificial intelligence for translational personalized neoantigen cancer vaccine development. Journal of Biomedical Science, 33(1), Article 87. https://doi.org/10.1186/s12929-026-01286-3

Image Credits: AI Generated

DOI: 10.1186/s12929-026-01286-3

Keywords: neoantigen cancer vaccine, artificial intelligence, personalized immunotherapy, HLA presentation, T-cell receptor, lipid nanoparticle, mRNA vaccine, precision oncology, immunopeptidomics, machine learning, cancer immunotherapy, translational oncology

Traditional antigen discovery has relied heavily on bulk omics. RNA sequencing of tumour biopsies identifies transcripts that are elevated in tumours relative to normal tissues, and mass spectrometry of immunopeptidomes identifies the peptides actually presented on human leukocyte antigen molecules. These approaches have been enormously productive, yielding the tumour-associated antigens that underpin therapeutic cancer vaccines, bispecific antibodies, and adoptive cell therapies. Yet they share a structural blind spot: they measure abundance, not arrangement. A transcript that is highly expressed in a tumour may also be expressed in a vital healthy tissue, and a peptide that appears abundant in a dissociated tumour sample may, in the intact tissue, be presented only on stromal cells rather than on the malignant compartment itself. Dissociation destroys the spatial relationships that the immune system, operating in intact tissue, would encounter.

Proximity labelling technologies emerged precisely to address this limitation. By fusing an engineered enzyme, such as a promiscuous biotin ligase or a peroxidase, to a protein of interest, researchers can covalently tag molecules that come within a few tens of nanometres of that protein in living cells. The tagged molecules can then be enriched and identified by mass spectrometry, producing a snapshot of the local molecular neighbourhood rather than the global composition of the cell. Applied to tumour biology, proximity labelling offers a way to ask what molecules physically congregate around a given marker protein, information that bulk profiling cannot provide. The microenvironment of a membrane protein, its partners, its neighbours in the plasma membrane, and the proteins trafficked alongside it, constitutes a layer of biological organisation that is invisible to standard transcriptomics and only partially accessible to conventional proteomics.

The challenge that arises once proximity data are generated is computational. A proximity labelling experiment produces a network-like structure: bait proteins, their tagged neighbours, the abundances of those neighbours, and the connections among them across conditions and cell states. Representing this as a simple list discards most of its meaning. Graph-structured data demand graph-aware analysis, and this is where modern graph learning methods become relevant. Graph neural networks and related architectures are designed to learn representations of nodes in a network that incorporate information from their local neighbourhoods, so that the identity of a protein is encoded not only by its own properties but by the company it keeps. In biological settings, this inductive bias is often exactly right: function in cellular systems is relational, and proteins that occupy similar network positions frequently share functional roles even when their sequences are unrelated.

Applying graph learning to proximity data in the tumour context creates an opportunity to formalise an intuition that immunologists have long held informally. An antigen that is safe to target therapeutically is not simply one that is absent from healthy tissue in a bulk measurement; it is one whose presentation is unlikely to occur on healthy cells under the conditions the immune system will actually encounter. Proximity provides a proxy for this contextual specificity. A candidate antigen that is consistently found in the immediate molecular neighbourhood of validated tumour markers, and that participates in the same spatial programmes as known malignant-surface proteins, carries more evidence of tumour association than a candidate identified by expression alone. Graph learning allows this evidence to be aggregated systematically across many candidates and many data modalities, rather than assessed one protein at a time by expert curation.

The therapeutic stakes of this problem are considerable. CAR T cell therapy, which engineers a patient’s T cells to recognise a surface antigen, has produced remarkable outcomes in haematological malignancies where a truly tumour-restricted target such as CD19 exists. Solid tumours have proven far more resistant, and a major reason is the absence of comparable target antigens. The antigens that are available on solid tumours, such as HER2, EGFR, or B7-H3, are frequently shared with essential healthy tissues, and targeting them produces on-target off-tumour toxicity that can be dose-limiting or fatal. The field has responded with a variety of engineering strategies, including logic-gated CARs that require two antigens to be present simultaneously, tunable affinity receptors that respond only to high antigen density, and adaptor systems that allow dosing control. All of these strategies depend on knowing which combinations of antigens are jointly specific to tumours, and this is a question about spatial co-organisation that proximity-guided approaches are well positioned to answer.

There is also a deeper immunological rationale for thinking in terms of proximity. The immune synapse itself is a proximity phenomenon: a T cell commits to killing only after sustained engagement with a target cell, integrating signals from many receptor-ligand interactions across the contact interface. Antigens that cluster together on the tumour surface may be recognised more effectively than antigens presented in isolation, because multivalent engagement strengthens T cell receptor signalling and can overcome the inhibitory signals that tumours deploy. Conversely, an antigen that is spatially segregated from co-stimulatory context may be immunologically silent even if it is abundant. Understanding the spatial organisation of tumour antigens therefore has implications not only for target selection but for predicting the quality of the immune response that targeting will provoke.

The tumour microenvironment adds further layers of complexity that proximity-aware methods are suited to capture. Tumours are not homogeneous masses of malignant cells; they are ecosystems containing fibroblasts, endothelial cells, macrophages, T cells, and extracellular matrix, all arranged in structured architectures that vary between patients and between regions of the same tumour. A candidate antigen expressed by tumour-associated fibroblasts, for example, might be attractive for stromal targeting strategies but inappropriate for direct tumour-cell killing. Distinguishing the malignant compartment from the reactive stromal compartment requires information about which proteins co-localise with which, and dissociated single-cell methods, while powerful, lose the tissue architecture that defines these compartments. Proximity labelling performed in intact systems, combined with graph-based inference, offers a route to recovering some of this architectural information from molecular data.

From a machine learning perspective, the tumour antigen problem illustrates a broader trend in computational biology: the shift from classification of individual entities to inference over relational structures. Early bioinformatics treated each gene or protein as an independent feature, and predictive models were built on expression vectors. The realisation that biological molecules operate in networks led to a family of methods that propagate information across interaction graphs, including network propagation algorithms, random walk approaches, and eventually graph neural networks. Each generation of methods has expanded the kinds of questions that can be asked. Where earlier approaches could ask whether a protein is connected to known disease genes, graph learning can ask whether a protein occupies a network position characteristic of disease-relevant molecules, a subtler and often more robust criterion. In the antigen discovery setting, this means the model can learn what the neighbourhood of a validated tumour antigen looks like and then score uncharacterised candidates by the similarity of their neighbourhoods.

Validation remains the essential counterweight to computational prediction, and the history of antigen discovery offers sobering lessons about candidates that looked compelling in silico but failed in vivo. A predicted proximity antigen must survive several successive tests: confirmation that the protein is genuinely presented on the tumour cell surface by human leukocyte antigen molecules, demonstration that healthy tissues lack comparable presentation, evidence that T cells capable of recognising the presented peptide exist in patients, and finally evidence that engaging those T cells produces tumour killing without tissue damage. Each of these tests is demanding, and attrition between stages is high. Computational methods that incorporate proximity information improve the front end of this pipeline by enriching the candidate list for molecules likely to pass the later stages, which is a meaningful advance even though no computational prediction can substitute for experimental validation.

The timing of this work within the broader technological landscape is notable. Spatial transcriptomics and spatial proteomics have matured rapidly, multiplexed imaging now routinely profiles dozens of proteins in intact tissue sections, and proximity labelling has been adapted to an expanding range of model systems. Meanwhile, immunopeptidomics has improved in sensitivity to the point where thousands of presented peptides can be catalogued from limited clinical material. The convergence of these technologies with graph-based machine learning creates conditions in which antigen discovery can become less dependent on serendipity. Historically, many successful tumour antigens were found through patient-specific approaches, such as isolating tumour-infiltrating lymphocytes and identifying their targets, which is powerful but slow and individualised. A systematic, proximity-informed discovery framework aims to generalise this process, producing catalogues of candidate antigens that can serve the broader population of patients.

There are also implications beyond T cell therapy. Antibody-drug conjugates require surface antigens with sufficient differential expression and internalisation behaviour, and the proximity context of an antigen can inform predictions about its trafficking and its accessibility to circulating antibodies. Bispecific molecules that bridge tumour cells and T cells depend on pairs of antigens whose joint expression pattern is tumour-restricted, and proximity data speak directly to co-localisation. Even vaccine design benefits, since peptide antigens that arise in the context of a tumour-specific molecular programme are more likely to elicit responses that discriminate tumour from self. In each of these therapeutic modalities, the fundamental question is the same: which molecular features distinguish the tumour cell surface in its intact context, and how confidently can that distinction be made for a given patient population.

As the field moves forward, the integration of proximity-guided graph learning into antigen discovery pipelines will likely be judged by a practical standard: whether the candidates it surfaces translate into therapies with wider therapeutic windows than those discovered by expression-based approaches alone. The conceptual contribution, however, may prove equally durable. By treating the tumour cell surface as a structured, relational system rather than a list of abundances, this line of work aligns the computational representation of tumour biology with the way the immune system itself reads that biology, through contact, context, and the molecular neighbourhoods in which every antigen is embedded.

Subject of Research: Proximity-guided graph learning reveals tumour-associated proximity antigens

Article Title: Proximity-guided graph learning reveals tumour-associated proximity antigens

Article References: Scandore, C., Malone, C. F., May, C. K., de Regt, A. K., Guernsey, J., Ma, H., Dephoure, N., Setter, B., Howell, R. A., Johnson, K. R., Farr, C. L., Romero, S., Vignale, L., Vittum, T., Dawson, E., Habtetsion, T., Nardi, F., Woodruff, B., Mathay, M., … Fadeyi, O. O. (2026). Proximity-guided graph learning reveals tumour-associated proximity antigens. Nature. https://doi.org/10.1038/s41586-026-11003-7

Image Credits: AI Generated

DOI: 10.1038/s41586-026-11003-7

Keywords: Proximity-guided, graph, learning, reveals, tumour-associated, proximity, antigens, scientific research