A Thirteen-Gene Immune Fingerprint That Predicts Survival — and Immunotherapy Success — in Lung Cancer
Lung adenocarcinoma, the most common form of lung cancer and the world’s deadliest malignancy, has acquired a new tool in the fight against its own unpredictability. Researchers at Tianjin Medical University Cancer Institute and Hospital in China have constructed and validated a prognostic signature built from thirteen immune infiltration-related genes — a molecular fingerprint that predicts how long patients with this disease are likely to survive and, remarkably, how likely they are to respond to immunotherapy. The study, published open access in the journal Cancer Cell International on 29 August 2026, was led by corresponding authors Ruifang Niu and Fei Zhang, with Jing Ren, Jiakun Gao and Shuhua Chen as co-first authors. Its central achievement is integration. Rather than offering yet another survival model, the 13-gene signature simultaneously forecasts prognosis, sorts tumors into molecular subtypes and anticipates therapeutic response, addressing one of oncology’s most stubborn problems: two patients with seemingly identical lung tumors can follow dramatically different courses, and clinicians have historically had little way to see the divergence coming.
The root of that unpredictability is heterogeneity. Lung adenocarcinoma, which arises from the mucus-producing glandular cells of the peripheral lung and accounts for the largest share of non-small cell lung cancers, is far from a single disease. Its clinical course ranges from indolent nodules cured by surgery to metastatic disease that resists every line of therapy. Two tumors of identical size and stage can harbor radically different immune landscapes, and those landscapes shape how the tumor grows, spreads and responds to treatment. Immune checkpoint inhibitors — the antibody drugs that release the molecular brakes holding T cells back — have transformed lung cancer therapy, yet only a minority of patients achieve durable benefit, and the tumor microenvironment features that govern response vary enormously from person to person. Because immune infiltration-related genes are closely intertwined with lung adenocarcinoma’s origin, progression and immunotherapy efficacy, the Tianjin team reasoned that these genes could yield a practical instrument capable of simultaneously dissecting a tumor’s immunobiological character and predicting its therapeutic response — a combination that existing clinical tools rarely deliver.
To build that instrument, the researchers mined RNA-sequencing data and clinical information for lung adenocarcinoma cohorts from The Cancer Genome Atlas (TCGA) and the Gene Expression Omnibus (GEO), two of the largest public archives of cancer genomics data. From the CIBERSORTx platform — a computational framework that deconvolves bulk RNA-sequencing data to estimate how many cells of each immune type are present inside an unsorted tumor sample — they retrieved 547 immune infiltration-related genes. Three successive statistical filters followed. Differential expression analysis flagged genes whose activity distinguished tumor tissue from healthy tissue. Cox proportional-hazards regression, the standard tool of survival analysis, then identified the subset of those genes whose expression levels were significantly linked to overall survival, assigning each gene a hazard ratio. Finally, least absolute shrinkage and selection operator (LASSO) regression — a machine-learning technique that penalizes model complexity by shrinking weak coefficients toward zero — distilled the panel to thirteen genes while guarding against the overfitting that undermines models built from thousands of candidates. The resulting signature was validated across multiple independent patient cohorts to confirm that its predictive power was not an artifact of any single dataset.
The finished product is deceptively simple: a single risk score, computed from the expression levels of thirteen genes, that sorts patients into high-risk and low-risk groups. Each patient’s tumor is assigned a score calibrated in a training cohort and then locked before testing. In multivariable Cox analysis, the score emerged as an independent prognostic factor, retaining its predictive power after accounting for other clinical variables and performing consistently across every validation cohort — a benchmark that many published genomic signatures fail to meet. The signature’s real payoff, however, lies in what the two risk groups turned out to represent. When the researchers compared molecular pathways, tumor stemness, the immune microenvironment, tumor mutational burden, immunotherapy responsiveness and drug sensitivity between groups, they found not two ends of a single continuum but two coherent immuno-biological phenotypes organized around fundamentally different biology. The high-risk phenotype was defined by runaway cell proliferation paired with immune evasion; the low-risk phenotype by vigorous immune infiltration and checkpoint expression. A single measurement computable from a standard tumor RNA-sequencing run captures a biological divide that conventional staging cannot see.
The high-risk group emerged as the fast, stealthy form of the disease. Its tumors showed enhanced activity in cell-cycle and DNA-replication pathways — the molecular engines of unrelenting division — along with increased stemness, quantified by an mRNA expression-based stemness index that measures how closely a tumor’s expression profile resembles that of embryonic stem cells. The metric, known as mRNAsi, condenses thousands of gene-expression values into one number that tracks how “stem-like” a tumor behaves, a quality generally associated with recurrence, therapy resistance and the self-renewing cell populations capable of regenerating a tumor after treatment. Paradoxically, these rapidly dividing tumors also carried a higher tumor mutational burden, the tally of mutations accumulated in a tumor’s genome. In many cancers a heavy mutational load works in immunotherapy’s favor, because mutations generate neoantigens — abnormal protein fragments that mark tumor cells for immune destruction. Here, however, the high mutational burden coexisted with an immunosuppressive microenvironment, producing tumors that are genetically loud yet immunologically quiet: bristling with potential targets, yet structurally equipped to hold attacking immune cells at bay.
The low-risk group told the opposite story. These tumors were densely infiltrated by immune cells, expressed higher levels of immune checkpoint genes — the very molecules targeted by checkpoint-inhibitor drugs — and showed higher predicted responsiveness to immunotherapy. To quantify that last property, the team applied two complementary computational gauges. The Tumor Immune Dysfunction and Exclusion (TIDE) algorithm estimates whether a tumor is likely to evade immunity through dysfunctional T cells or through physical exclusion of immune cells from the tumor mass, with lower predicted dysfunction generally signaling greater benefit from checkpoint blockade. The Immunophenoscore, derived from The Cancer Immunome Atlas, integrates antigen presentation, immunomodulatory signaling and other parameters into a composite measure of a tumor’s intrinsic immunogenicity. Both metrics serve as computational stand-ins for the question every oncologist faces: once the immune system is unleashed, will it actually attack this tumor? By both measures, low-risk tumors profiled as the stronger candidates for immunotherapy, while high-risk tumors registered as more resistant.
The model’s reach extends into pharmacology as well. By cross-referencing the two risk groups with the Genomics of Drug Sensitivity in Cancer database (GDSC2), the team estimated the half-maximal inhibitory concentration — IC50, the drug concentration that halves the viability of cancer cells in laboratory assays — for a panel of candidate therapeutics. The risk groups differed in their predicted drug sensitivities, suggesting that the 13-gene signature could help steer not only immunotherapy decisions but also the selection of conventional systemic treatments. Such predictions cannot prescribe a regimen on their own, but they flag candidates for laboratory follow-up and help prioritize combinations for each molecular subtype. Read as a clinical instrument, the signature works as a therapeutic stratification engine: a patient whose tumor falls into the immune-infiltrated, checkpoint-rich low-risk category might be steered toward immunotherapy, whereas a high-risk, stem-like, proliferation-driven profile could argue for adjuvant intensification or for combination strategies designed to convert a cold tumor into a hot one by drawing immune cells into tumors that currently exclude them.
Crucially, the investigators did not allow the model to remain purely computational. One of its thirteen constituents, SKA1, was carried into the wet laboratory for functional testing. SKA1 encodes the spindle and kinetochore-associated protein 1, part of the molecular machinery that physically tethers chromosomes to the mitotic spindle during cell division — a role that makes it a logical suspect in a cancer defined by unchecked proliferation. Using small interfering RNA to silence SKA1 in lung adenocarcinoma cells, and verifying protein changes by Western blotting with SDS-polyacrylamide gel electrophoresis and polyvinylidene fluoride membranes, the researchers measured what happens when the gene is removed. The answer was unambiguous: SKA1 behaves as an oncogenic driver in lung adenocarcinoma, significantly enhancing the proliferation, migration and invasion of cancer cells. The finding anchors the computational signature in experimental biology, showing that at least one of its thirteen genes is not merely correlated with aggressive disease but functionally implicated in driving it — and hinting that the remaining genes may encode further vulnerabilities worth targeting.
As with any model built from retrospective datasets, caveats remain. The signature must now survive prospective validation in clinical trials, and converting thirteen genes into a routine diagnostic will demand standardized measurement, regulatory review and evidence that treatment guided by the score genuinely improves survival; the journal is likewise releasing the peer-reviewed article early, with a final version of record to follow. Even so, the scope of the study is unusual. In one integrated framework — supported by grants from the National Natural Science Foundation of China and the Tianjin Key Medical Discipline Construction Project — the team delivered a validated prognostic model, revealed two distinct immuno-biological phenotypes of lung adenocarcinoma, linked each phenotype to predicted immunotherapy response and drug sensitivity, and confirmed a key oncogenic gene at the bench. For a disease that still claims more lives than any other cancer, the implications run in two directions at once: the immune microenvironment encodes both prognosis and therapeutic opportunity, and thirteen genes may be enough to read it.
Subject of Research: Development and validation of a 13-gene immune infiltration-based prognostic signature for survival prediction, molecular subtyping, and immunotherapy-response stratification in lung adenocarcinoma (LUAD).
Subject of Research: Cancer
Article Title: Immune infiltration-based prognostic signature for molecular subtyping and therapeutic stratification in lung adenocarcinoma
Article References: Ren, J., Gao, J., Chen, S., Liu, L., Zhang, H., Song, L., Guo, H., Wang, Z., Zhou, Y., Cui, Y., Niu, R., & Zhang, F. (2026). Immune infiltration-based prognostic signature for molecular subtyping and therapeutic stratification in lung adenocarcinoma. Cancer Cell International. https://doi.org/10.1186/s12935-026-04456-3
Image Credits: AI Generated
DOI: 10.1186/s12935-026-04456-3
Keywords: Lung adenocarcinoma, Prognostic model, Immune infiltration, Immunotherapy, Tumor microenvironment, Molecular subtyping, Tumor mutational burden, LASSO regression, SKA1, Immune checkpoint inhibitors
Cite Scienmag News
APA
MLA
Chicago
Nathaniel Bowman. (August 30, 2026). Immune signature predicts prognosis and guides therapy in lung adenocarcinoma. Scienmag. https://scienmag.com/immune-signature-predicts-prognosis-and-guides-therapy-in-lung-adenocarcinoma/
Nathaniel Bowman. “Immune signature predicts prognosis and guides therapy in lung adenocarcinoma.” Scienmag, 30 August 2026, https://scienmag.com/immune-signature-predicts-prognosis-and-guides-therapy-in-lung-adenocarcinoma/. Accessed 30 August 2026.
Nathaniel Bowman. “Immune signature predicts prognosis and guides therapy in lung adenocarcinoma.” Scienmag. August 30, 2026. https://scienmag.com/immune-signature-predicts-prognosis-and-guides-therapy-in-lung-adenocarcinoma/
Copy citation
Download RIS
Tags: and address the challenges posed by tumor heterogeneity in lung adenocarcinoma.and immunotherapy responsiveness. This advancement aims to personalize therapycomplicating prognosis and treatment strategies. The newly developed 13-gene immune signature provides a comprehensive approach by integrating immune infiltration profiles to predict patient survivalenhancing personalized treatment strategies.exhibits high cellular and molecular heterogeneityexhibits significant molecular and immune heterogeneity. The newly developed immune fingerprint leverages this heterogeneity to provide a comprehensive prognostic and predictive toolimprove outcome predictionslung tissuetumor subtypes
