——Getting oriented on spatial biomarkers starts with a single distinction that explains almost everything else in this fast-growing field: a conventional biomarker answers what is present, while a spatial biomarker answers what is present, where within the tissue, and next to what other cells or structures. That third piece of information, location and neighborhood, is not a minor refinement. Across the studies this primer surveys, it is frequently the specific detail that separates a biomarker that predicts outcome from one that does not.
Key takeaways
|
Defining a spatial biomarker
A spatial biomarker is a predictive or diagnostic signature defined by the spatial arrangement, density, or co-occurrence of molecular or cellular features within tissue, rather than by the abundance of a single molecule measured without regard to position. The defining property is that the signature cannot be reduced to one number, such as percent positive cells, without losing the information that makes it predictive.
That definition is broader than it might first appear, because it covers several genuinely different kinds of measurement. A spatial biomarker might describe how many of a given cell type are present in a region, how close two cell types sit to one another, whether two specific proteins are physically interacting, which markers co-occur on the same cell, or what composition of surrounding cell types characterizes a given tissue neighborhood. All of these are spatial in the sense that none of them survives tissue dissociation, and all of them require a technology capable of preserving position during measurement.
Spatial vs. conventional biomarkers
The clearest way to see the difference is to compare what each approach can and cannot answer about the same tissue sample.
Question | Conventional biomarker | Spatial biomarker |
Is a protein present, and how much | Answers this directly and reliably | Can answer this too, but treats it as only one input among several |
Where in the tissue is it located | Cannot answer; this information is typically lost or averaged away | Answers this directly, since position is preserved during measurement |
Is it physically near or interacting with another feature | Cannot answer at all | Specifically designed to answer this |
Does the surrounding tissue composition matter | Not captured | Captured directly as neighborhood or niche composition |
Table 1. What conventional and spatial biomarkers can each answer about the same tissue sample. A spatial biomarker does not replace a conventional one; it adds dimensions a conventional measurement cannot capture at all.
Types of spatial features
Five distinct categories of spatial feature recur across the current literature, and distinguishing between them matters because each requires different data, different analysis, and answers a different biological question.
- Density and abundance. How many of a given cell type or how much of a given marker is present within a defined tissue region, the spatial analog of a conventional biomarker but assessed region by region rather than across a whole homogenized sample.
- Direct molecular interaction. Whether two specific proteins, such as a checkpoint receptor and its ligand, are physically engaging one another, confirmed through techniques that detect direct binding rather than inferring it from co-location. Our coverage in Spatial Biology in Oncology: Decoding the Tumor Microenvironment details a head and neck cancer study using exactly this kind of direct-interaction evidence.
- Co-expression. Which markers are simultaneously present on the same individual cells, rather than merely present somewhere in the same tissue sample. A four-marker exhaustion signature covered in Spatial Biology in Immuno-Oncology: Targets, Signatures, and Combinations is a concrete example of this feature type.
- Proximity and distance. How physically close one cell type sits to another, often measured as a nearest-neighbor distance or as a count of one cell type within a fixed radius of another. Predicting Response to Immunotherapy With Spatial Signatures surveys several studies built entirely around this feature type.
- Neighborhood or niche composition. What mixture of surrounding cell types characterizes a specific tissue region, capturing local microenvironment context rather than any single cell-to-cell relationship. Structural exclusion mechanisms covered in Mapping the Tumor Immune Microenvironment for Immuno-Oncology and resistance niches covered in Spatial Biology in Oncology: Decoding the Tumor Microenvironment both illustrate this feature type.
Why context adds predictive power
The motivation for pursuing spatial biomarkers at all is not merely that more data is available. It is a specific, stated dissatisfaction with how well conventional biomarkers currently perform.
A 2024 review evaluating biomarkers for predicting response to immune checkpoint inhibition states this directly: current predictive biomarkers for checkpoint inhibitor response remain sub-optimal, and the review specifically investigates whether assessing the spatial distribution and interaction of cellular components in clinical samples could identify superior biomarkers of response. That framing, an explicit acknowledgment that the conventional approach has a real performance ceiling, is the honest starting point for why this entire field exists, rather than assuming spatial methods are self-evidently better without stating what problem they are meant to solve.
The mechanistic reason context adds predictive power follows directly from the taxonomy above: a molecule’s abundance alone cannot capture whether it is physically positioned to do the biological work a therapy depends on. A checkpoint protein far from any T cell, a resistant cell isolated from any protective niche, or a marker expressed on cells scattered randomly rather than co-expressed together, all represent biologically different situations that an abundance-only measurement reports identically.
Examples in development
A study spanning three distinct multiplex imaging technologies illustrates several of these feature types operating together, and its cross-platform design is itself methodologically important for a primer aimed at drug developers.
One study, three platforms, multiple feature types at onceResearchers analyzed 102 breast cancer patients using cyclic immunofluorescence data they generated directly, combined with publicly available imaging mass cytometry and multiplex ion beam imaging data from additional cohorts, specifically to test whether spatial biomarkers discovered on one platform reproduce on others. Cellular abundance and proximity-based biomarkers with prognostic value were identified and shown to be reproducible across all three platforms, including lymphocyte infiltration independently associated with longer survival in triple-negative and high-proliferation breast tumors. In triple-negative breast cancer specifically, the study found macrophage proximity to tumor cells and B cell proximity to T cells were greater in good-prognosis tumors, a proximity-type feature, while tumor neighborhoods enriched in vimentin-positive fibroblasts were associated with poor prognosis, a neighborhood-composition-type feature. One study, three feature types, three imaging platforms: that combination is exactly the kind of cross-platform reproducibility a drug developer should look for before treating any single spatial finding as generalizable. |
How a spatial biomarker like these moves from a research finding toward validated clinical use, including discovery methodology specifically, is developed next in Spatial Biology in Biomarker Discovery. The regulatory and commercial path from a validated spatial signature to an approved companion diagnostic is covered in full in Spatial Biomarkers and Companion Diagnostics: The Next Frontier. For where biomarkers sit within the full spatial biology pipeline, see Spatial Biology in Drug Discovery: From Target Discovery to Translational Medicine.
This article was produced under Drug Discovery News’s AI editorial policies.
