How is spatial biology used in oncology?

Spatial biology maps where tumor, immune, and stromal cells sit relative to one another within a tumor, revealing interactions and resistance mechanisms that bulk or dissociated single-cell data cannot detect. Recent applications include directly mapping checkpoint protein interactions, identifying specific immunosuppressive cell neighborhoods, and correlating spatial protein expression with clinical outcomes such as time to relapse.

What is the tumor microenvironment in drug discovery?

The tumor microenvironment is the ecosystem surrounding malignant cells, including immune cells, stromal cells, blood vessels, and extracellular matrix, that collectively influence tumor growth and treatment response. In drug discovery, it matters because a mechanistically sound therapy can still fail if the surrounding tissue architecture physically excludes the immune cells or blocks the interactions the therapy depends on.

Why is the TME important for cancer drugs?

Because treatment response often depends on tissue architecture as much as on the tumor cells themselves. A 2024 study found that a specific spatial neighborhood of immunosuppressive macrophages and stem-like tumor cells drove treatment resistance in hepatocellular carcinoma, and that a combination therapy targeting the specific mechanism within that neighborhood eliminated the resistant cells in mouse models.

Spatial biology in oncology and the TME

Highlights in this article

The TME as a drug discovery problem
Mapping tumor-immune interactions
Resistance and immune-evasion niches
Spatial targets in the TME
From map to drug program

Understanding tumor microenvironment drug discovery increasingly means understanding neighborhoods rather than averages. A tumor is not simply malignant cells plus whatever else happens to be nearby; it is an ecosystem where which cells sit next to which other cells, and what those neighbors do to each other can decide whether a targeted therapy or checkpoint inhibitor actually works in that specific patient.


Key takeaways Spatial biology has been adopted faster in oncology than in almost any other research area, because tumor microenvironment biology is inherently a question of physical position. A 2024 study of hepatocellular carcinoma minimal residual disease identified a specific macrophage-tumor cell spatial neighborhood, then tested and validated a combination therapy targeting the mechanism driving it. A separate hepatocellular carcinoma study named the exact molecular mechanism, interleukin-1 beta secretion, by which specific macrophages suppress immune response, validated across eight patient cohorts. Spatially resolved protein expression in prostate cancer tumor and stroma has been directly correlated with time to biochemical relapse in a cohort of 97 patients. Direct protein-protein interaction mapping, not just co-expression, is emerging as a stronger form of spatial evidence for predicting immunotherapy response.

Key takeaways

Spatial biology has been adopted faster in oncology than in almost any other research area, because tumor microenvironment biology is inherently a question of physical position.
A 2024 study of hepatocellular carcinoma minimal residual disease identified a specific macrophage-tumor cell spatial neighborhood, then tested and validated a combination therapy targeting the mechanism driving it.
A separate hepatocellular carcinoma study named the exact molecular mechanism, interleukin-1 beta secretion, by which specific macrophages suppress immune response, validated across eight patient cohorts.
Spatially resolved protein expression in prostate cancer tumor and stroma has been directly correlated with time to biochemical relapse in a cohort of 97 patients.
Direct protein-protein interaction mapping, not just co-expression, is emerging as a stronger form of spatial evidence for predicting immunotherapy response.

The TME as a drug discovery problem

The tumor microenvironment, meaning the malignant cells plus the immune cells, stromal cells, blood vessels, and extracellular matrix that surround and interact with them, has moved from a peripheral consideration in oncology research to a central one, and the shift has been driven specifically by the recognition that treatment response depends on that surrounding architecture as much as on the tumor cells themselves.

Cancer research has adopted spatial biology faster than almost any other field, largely because tumor microenvironment biology is inherently spatial. Whether immune cells can physically reach and infiltrate a tumor, and what signals surrounding stromal cells send in response, are questions dissociated sequencing simply cannot answer, and spatial platforms have already been used to map how cancer and immune cell communication patterns relate to a patient’s response to immunotherapy. Our colleagues at Technology Networks survey this ground, and several adjacent application domains, in Applications of spatial biology: From tumor microenvironment to brain mapping, which frames the TME as one instance of a broader principle: where a cell sits determines what signals it receives, and what behavior it exhibits, a principle spatial methods make directly measurable rather than inferred.

For a drug discovery program specifically, that principle has a blunt practical consequence. A candidate immunotherapy can be mechanistically sound and still fail in a specific patient if that patient’s tumor architecture physically excludes the immune cells the drug depends on, or if a specific stromal or myeloid population sits positioned to neutralize the drug’s effect before it can act. Understanding the TME spatially is therefore not an academic refinement; it is increasingly a prerequisite for understanding why an oncology drug worked in some patients and not others.

Mapping tumor-immune interactions

The most direct application of spatial biology to the TME is mapping where immune cells sit relative to tumor cells and to each other, and recent work has pushed this from co-expression toward direct physical interaction, which is a meaningfully stronger form of evidence.

A study using in situ proximity ligation assay, a technique that detects direct protein-protein interaction rather than inferring it from the two proteins simply being present in the same region, mapped PD-1/PD-L1 interactions directly in head and neck squamous cell carcinoma tissue collected before patients received checkpoint inhibitor therapy. The mapping identified macrophage-tumor barriers associated with immunotherapy response, work the authors situate within spatial proteomics’ recognition as Nature’s Method of the Year for 2024.


Why direct interaction evidence is a meaningful upgrade Two proteins detected in the same tissue region are not necessarily in contact with each other. A macrophage expressing PD-L1 and a T cell expressing PD-1, both present in a given field of view, could be too far apart within that region to actually engage the checkpoint interaction the co-expression pattern implies. Proximity ligation assay and comparable techniques close that specific gap, confirming physical interaction rather than merely spatial co-occurrence. That distinction matters directly for a biomarker development program. A co-expression-based spatial signature and a direct-interaction-based spatial signature are different claims with potentially different predictive value, and a program building a companion diagnostic around checkpoint biology should know which kind of evidence its assay actually measures.

Why direct interaction evidence is a meaningful upgrade

Two proteins detected in the same tissue region are not necessarily in contact with each other. A macrophage expressing PD-L1 and a T cell expressing PD-1, both present in a given field of view, could be too far apart within that region to actually engage the checkpoint interaction the co-expression pattern implies. Proximity ligation assay and comparable techniques close that specific gap, confirming physical interaction rather than merely spatial co-occurrence.

That distinction matters directly for a biomarker development program. A co-expression-based spatial signature and a direct-interaction-based spatial signature are different claims with potentially different predictive value, and a program building a companion diagnostic around checkpoint biology should know which kind of evidence its assay actually measures.

Resistance and immune-evasion niches

Spatial data has moved beyond identifying that resistance or immune evasion happens somewhere in a tumor, toward naming which specific cells drive it and by what molecular mechanism, which is a considerably more actionable form of evidence for a drug discovery program.

A 2024 Nature Cancer study profiled 401 hepatocellular carcinoma samples using co-detection by indexing across 36 biomarkers, identifying vimentin-high macrophages whose spatial co-occurrence with regulatory T cells was validated across eight independent patient cohorts and shown to promote tumor progression. Functional follow-up work went a step further, demonstrating that these macrophages enhance the immunosuppressive activity of regulatory T cells specifically by increasing secretion of interleukin-1 beta. That is a named molecular mechanism behind a spatial pattern, not simply a description of the pattern itself, which is what makes the finding testable and, in principle, targetable.

A separate, related niche has been characterized in hepatocellular carcinoma minimal residual disease specifically, the tumor cells that persist after treatment and later cause relapse. That work is developed in full in the following section, since it is this cluster’s clearest example of a spatial niche finding translating directly into a tested therapy.

Spatial targets in the TME

The clearest demonstration in current oncology literature of a spatial finding moving all the way from observation to tested intervention comes from a 2024 study of hepatocellular carcinoma minimal residual disease, the population of tumor cells that survives treatment and later drives recurrence.


From a spatial neighborhood to a validated combination therapy Using single-cell, high-plex cytometric imaging on human tissue collected after chemoembolization, alongside a transgenic mouse model, researchers identified a specific spatial neighborhood in which PD-L1-positive, M2-like macrophages interact directly with stem-like tumor cells. That neighborhood correlated with CD8-positive T cell exhaustion and with poor survival. Spatial transcriptomics of the residual disease then showed that macrophage-derived transforming growth factor beta mediates the persistence of the stem-like tumor cells within that neighborhood, naming a specific molecular driver rather than only describing the correlation. The researchers then tested the logical intervention directly: combined blockade of PD-L1 and TGF-beta. In two separate mouse models, that combination excluded the immunosuppressive macrophages, recruited activated CD8-positive T cells, and eliminated the residual stem-like tumor cells. That is the complete pipeline this cluster describes, realized in a single body of work: a spatial observation identifies a specific cellular neighborhood, mechanism work names the molecular driver within it, and a therapy targeting that mechanism is tested and shown to work.

From a spatial neighborhood to a validated combination therapy

Using single-cell, high-plex cytometric imaging on human tissue collected after chemoembolization, alongside a transgenic mouse model, researchers identified a specific spatial neighborhood in which PD-L1-positive, M2-like macrophages interact directly with stem-like tumor cells. That neighborhood correlated with CD8-positive T cell exhaustion and with poor survival. Spatial transcriptomics of the residual disease then showed that macrophage-derived transforming growth factor beta mediates the persistence of the stem-like tumor cells within that neighborhood, naming a specific molecular driver rather than only describing the correlation.

The researchers then tested the logical intervention directly: combined blockade of PD-L1 and TGF-beta. In two separate mouse models, that combination excluded the immunosuppressive macrophages, recruited activated CD8-positive T cells, and eliminated the residual stem-like tumor cells. That is the complete pipeline this cluster describes, realized in a single body of work: a spatial observation identifies a specific cellular neighborhood, mechanism work names the molecular driver within it, and a therapy targeting that mechanism is tested and shown to work.

Clinical correlation, as distinct from mechanistic validation, is a different and complementary form of evidence, and a prostate cancer study supplies it directly. Researchers applied digital spatial profiling to assess 58 proteins and 1,825 RNA transcripts in archived formalin-fixed, paraffin-embedded tissue, then validated selected protein findings in a tissue microarray of 1,547 cores from 97 patients, testing whether OX40L, CTLA4 and CD11c expression in tumor and tumor-adjacent stroma correlated with time to biochemical relapse. Tying spatially resolved protein expression to an actual clinical outcome measure, in a cohort of that size, is a meaningfully stronger form of evidence for a biomarker program than a mechanistic finding in cell culture or mouse models alone, even though it does not by itself establish the causal mechanism the hepatocellular carcinoma work above does.

From map to drug program

Turning findings of this kind into an actual drug discovery program means asking a specific sequence of questions rather than treating every spatial correlation as equally actionable.

Is the spatial pattern mechanistically explained, or only correlated? The hepatocellular carcinoma minimal residual disease work names TGF-beta and interleukin-1 beta as specific mechanisms; the prostate cancer correlation with biochemical relapse, while real and clinically meaningful, has not yet been mechanistically dissected to the same degree in the source discussed here.
Does the mechanism suggest an existing drug class, or a novel target? PD-L1 and TGF-beta blockade both used approved or clinically advanced drug classes in the hepatocellular carcinoma example, which shortened the path from spatial finding to testable combination considerably compared with pursuing an entirely novel target.
Has the finding been validated across independent patient cohorts? The vimentin-high macrophage finding was checked across eight independent cohorts before its mechanism was pursued functionally, which is the same reproducibility discipline argued for elsewhere in this cluster’s target discovery coverage.
Does the evidence type match the claim being made? A direct-interaction spatial signature, a co-expression spatial signature, and a clinical-outcome correlation are three different strengths of evidence, and a drug program should know explicitly which one it is building on before setting expectations for how predictive that evidence will be.

Each of the topics introduced in this overview, mapping the tumor immune microenvironment specifically, spatial transcriptomics methods in cancer research, tumor heterogeneity and clonal architecture, immuno-oncology targets and combinations, and predicting immunotherapy response with spatial signatures, is developed in full in its own dedicated guide within this section: Mapping the Tumor Immune Microenvironment for Immuno-Oncology, Spatial Transcriptomics in Cancer Research and Drug Development, Tumor Heterogeneity and Spatial Clonal Architecture, Spatial Biology in Immuno-Oncology: Targets, Signatures, and Combinations, and Predicting Response to Immunotherapy With Spatial Signatures.

For where oncology and the tumor microenvironment sit within the full spatial biology pipeline, from target discovery through biomarkers to clinical translation, see Spatial Biology in Drug Discovery: From Target Discovery to Translational Medicine.

This article was produced under Drug Discovery News’s editorial policies.