New Spatial Index Measures the Immune Barrier That Shields Tumours From Attack

One of the most frustrating puzzles in cancer immunology is why a tumour can be swarming with immune cells and still grow unchecked. The answer, increasingly, lies not in how many immune cells are present but in where they are and what stands between them and their target. A study published in the Journal of Translational Medicine by Yuanyuan Liu and Minjuan Xu of Ganzhou People’s Hospital in China tackles this problem head-on, introducing a computational measure designed to capture something that conventional immune-cell counting cannot: the suppressive barrier that holds immune cells at the tumour’s edge and prevents them from penetrating the tumour core.

The researchers call their measure the Spatial Immune Barrier Index, or SIBI. It is built on spatial transcriptomics, a technology that maps gene expression across intact tissue sections while preserving the physical relationships between cells. Rather than summarising the immune infiltrate as a single abundance figure, SIBI combines two distinct pieces of spatial information. The first is an immune-presence gate, which asks whether immune cells are actually detectable at the tumour periphery. The second is a minimax barrier term, computed over the possible paths leading from the tumour core to the tumour’s exterior, which quantifies the strength of the suppressive signals that any infiltrating immune cell would have to cross on its way in.

The logic behind this construction reflects a tripartite view of tumour immune geography that oncologists have long recognised but struggled to quantify. In immune-desert tumours, few immune cells are present anywhere, and the immune system appears never to have engaged with the malignancy. In immune-infiltrated tumours, immune cells penetrate deep into the tumour parenchyma and can engage cancer cells directly. In immune-excluded tumours, the third and most ambiguous category, immune cells are abundant but accumulate at the tumour margin, held at bay by a suppressive microenvironment. Abundance-based summaries, the standard tools of pathology and bulk sequencing, cannot distinguish the excluded state from the infiltrated one, because both register as immune-positive. SIBI was designed specifically to break that ambiguity.

To test the index, the authors applied it to two independent cancer types profiled by spatial transcriptomics: high-grade serous ovarian carcinoma, an aggressive malignancy that accounts for the majority of ovarian cancer deaths, and head and neck squamous cell carcinoma, a tumour type in which immune exclusion is a well-documented barrier to immunotherapy. Across these two spatial cohorts, SIBI separated samples with favourable immune characteristics from those with unfavourable ones more effectively than immune abundance alone. The area under the curve, a standard measure of classification performance, reached 0.90 for SIBI compared with 0.71 for abundance-based metrics, a substantial difference in a field where small margins often decide whether a biomarker is pursued.

Robustness was a central concern in the study’s design. Any index built from spatial data involves choices: which genes define immune presence, how the suppressive programme is scored, how paths through the tissue are enumerated. The authors report that SIBI’s ranking of samples was stable across the range of choices made in its construction, meaning the measure does not hinge on arbitrary parameter settings. They also benchmarked SIBI against ten established and classical spatial analysis methods, and found that it outperformed all of them in separating outcome groups within the discovery cohorts. That comparison matters because spatial transcriptomics has produced a crowded field of scoring schemes, and few have been tested against one another on the same data.

Perhaps the most striking claim in the paper concerns reproducibility beyond the discovery setting. The distribution of SIBI values, the authors report, reproduces across seven independent external cohorts spanning two profiling platforms and four cancer types. In an era when many proposed biomarkers fail to replicate across laboratories and technologies, that kind of cross-platform consistency is notable. The external cohorts serve a specific purpose in the study’s logic: they establish that the index itself is reproducible, not that it independently predicts clinical outcomes. The authors are explicit about this distinction, and it shapes how the findings should be read.

Prognostic evidence comes instead from an independent bulk ovarian-cancer cohort with survival data. Because bulk tissue lacks the spatial resolution needed to compute SIBI directly, the authors tested the underlying biology rather than the index itself, asking whether the directional suppression programme that drives the barrier term carried prognostic information. It did: in that cohort, the programme retained significant prognostic value for overall survival. This suggests that the biology SIBI captures, the coordinated suppressive activity at the tumour boundary, is not merely a spatial artefact but a clinically meaningful feature of the tumour microenvironment that leaves a detectable signature even in non-spatial data.

An important conceptual contribution of the work is the separation of two kinds of barriers that are often conflated in discussions of immune exclusion. A physical barrier, built largely from extracellular matrix and abnormal vasculature, can mechanically obstruct immune-cell migration into the tumour. A functional barrier, by contrast, consists of active immunosuppressive signalling, such as checkpoint ligands, suppressive cytokines and regulatory cell populations, that disables immune cells even when they reach the tumour margin. The authors present evidence that SIBI captures the functional, immunosuppressive barrier specifically, distinct from the physical extracellular-matrix barrier. That distinction has therapeutic implications, because the two barrier types would plausibly respond to different interventions, from matrix-modifying agents to checkpoint inhibitors and cytokine blockade.

The clinical implications, if the findings hold up, could be considerable. Immune-excluded tumours are a major reason why checkpoint immunotherapy benefits only a fraction of patients: drugs that release the brakes on T cells work best when those T cells are in contact with tumour cells, and they accomplish little when the effector cells are stranded outside a suppressive perimeter. A reproducible, interpretable measure of exclusion could help identify patients who need combination strategies aimed at dismantling the barrier, rather than therapies that assume immune cells are already in position. It could also serve as a pharmacodynamic readout in trials of drugs designed to remodel the tumour microenvironment, showing whether a candidate agent actually weakens the barrier it targets.

The authors are careful to position SIBI as an exploratory research index rather than a validated clinical classifier, and the caveats they attach are worth taking seriously. The external cohorts demonstrate reproducibility of the index, not independent outcome validation, and the prognostic claims are anchored to the discovery and survival cohorts rather than to fresh clinical data. The clear next step, as the paper states, is prospective multi-centre validation, in which SIBI is computed on spatially profiled tumours from patients whose treatment courses and outcomes are followed forward in time. Until that work is done, SIBI should be understood as a promising measurement tool for researchers, one that converts a fuzzy pathological intuition, the sense that immune cells are being kept out, into a number that can be compared across tumours, platforms and studies. If it survives prospective testing, it could give oncologists a way to see not just whether the immune system has arrived at a tumour, but whether it has been allowed in.

Subject of Research: A spatial transcriptomic index quantifying immune exclusion barriers in the tumour microenvironment

Article Title: A spatial index of immune barriers in the tumour microenvironment

Article References: Liu, Y., & Xu, M. (2026). A spatial index of immune barriers in the tumour microenvironment. Journal of Translational Medicine. https://doi.org/10.1186/s12967-026-09024-x

Image Credits: AI Generated

DOI: 10.1186/s12967-026-09024-x

Keywords: spatial transcriptomics, tumour microenvironment, immune exclusion, immunotherapy response, ovarian cancer, head and neck squamous cell carcinoma, tumour immunology, immune evasion, biomarker, cancer microenvironment, prognosis, computational biology