Few technologies have moved as quickly from academic curiosity to pipeline tool as spatial biology in drug discovery. In 2024, Nature Methods named spatial proteomics its Method of the Year, citing its impact on how researchers understand tissue organization, cell-cell interaction and the tumor microenvironment, and noting the acceleration of atlas-scale consortium projects that are mapping human tissue in unprecedented detail. That recognition marks a field that has moved past proof of concept and into the tools drug discovery teams use to find targets, understand mechanism and build biomarkers.
Key takeaways
|
What spatial biology brings to drug discovery
A tumor biopsy dissolved into a tube for sequencing gives up its cellular composition and loses its architecture in the same step. Every gene expression value, every protein abundance, becomes an average across whatever mixture of cell types happened to be in that tube, with no record of which cells sat next to which, or how far a signaling molecule had to travel to reach its target.
Spatial biology keeps that architecture intact. Spatial transcriptomics, spatial proteomics, and related imaging methods measure gene expression or protein abundance while preserving the exact tissue coordinates of each measurement, so a researcher can ask not just what is present but where, and next to what. As one industry analysis of the field put it, traditional target identification relies heavily on bulk transcriptomics or proteomics, which can obscure the cell-type-specific or spatially restricted expression of potential drug targets, a limitation spatial methods are specifically designed to remove.
—Nature Methods, Method of the Year 2024 collection
"Spatial proteomics has transformed cancer research by providing unparalleled insights into the microenvironmental landscape of tumors."
That recognition was not an isolated editorial choice. Nature Methods’ 2024 selection specifically credited spatial proteomics with advancing understanding of tissue organization, cell-to-cell interaction and spatially coordinated mechanisms behind antitumor immune responses, and pointed to large-scale efforts including the Human BioMolecular Atlas Program and the Human Tumor Atlas Network as evidence the field had moved from single studies to shared infrastructure. Those consortium atlases matter for a drug discovery audience specifically because they are building the reference maps against which any new spatial dataset can be compared, the same way reference genomes underpin modern genomics.
Three capabilities are what pharma teams are actually buying when they adopt spatial methods, and they map directly onto this cluster’s structure.
- Cell-type-resolved target expression. Knowing which specific cell type expresses a candidate target, and where that cell type sits relative to disease-relevant structures, rather than an average across a mixed tissue sample.
- Microenvironment context. Multiplexed imaging platforms can now detect more than 100 protein targets in a single tissue section, which is what makes characterizing the full immune and stromal composition of a tumor practical rather than theoretical.
- A new biomarker class. Spatial phenotypic signatures, defined by cell density and interaction patterns rather than expression level alone, are a genuinely new type of readout that existing single-marker diagnostics cannot produce.
This guide works through each of those in turn: target discovery in tissue context, the tumor microenvironment specifically, spatial biomarkers and companion diagnostics, and the translational path into clinical use, closing with where the field is headed next. A companion overview of the broader field’s momentum is available in our earlier piece, The rise of spatial omics.
Target discovery in the tissue context
Identifying a drug target from bulk tissue data means identifying a gene or protein that is different, on average, between diseased and healthy samples. That average can be misleading in two specific ways spatial data corrects.
- A target expressed by a rare but critical cell population can be diluted below detection in bulk data. If the relevant cell type is a small fraction of the sample, its expression signature is averaged against everything else and can disappear entirely.
- A target’s relevance can depend on its neighbors. The same protein expressed by a tumor cell adjacent to an exhausted T cell may represent a fundamentally different therapeutic opportunity than the identical protein expressed by a tumor cell in a region with no immune infiltration at all, a distinction bulk sequencing cannot make because it has already discarded position.
Atlas-scale consortium projects are building the infrastructure that makes this kind of discovery tractable at scale rather than one study at a time. The Human Tumor Atlas Network, one of the programs credited in Nature Methods’ 2024 recognition of the field, is producing shared, spatially resolved reference datasets specifically so that a new candidate target can be checked against an existing map of tumor cellular composition before committing resources to validate it experimentally.
Two computational developments from that same body of work are worth knowing about because they represent the kind of tooling problem target discovery teams are likely to encounter directly. CyLinter is a quality-control tool built to clean up the artifacts that accumulate in highly multiplexed images before they are analyzed, addressing a problem specific to spatial data: a segmentation or staining error in a busy image is much harder to catch than the same error in a simpler dataset. CalicoST is an algorithm that infers copy number changes and reconstructs how a tumor’s genetic makeup varies across a section directly from spatially resolved transcriptomic data, connecting genomic evolution to physical location within the tumor rather than treating both as separate analyses.
Discovery question | What bulk data tells you | What spatial data adds |
Is this target expressed in the disease tissue | An average expression level across the whole sample | Which specific cell type expresses it, and at what frequency |
Is the target druggable in context | Presence or absence of the molecule | Whether the cells expressing it sit near the structures a therapy needs to reach |
Does the target matter for this patient population | A population-level association | Whether the spatial pattern reproduces across patients or reflects one unusual sample |
Is a candidate genuinely novel | Differential expression versus a reference dataset | Comparison against atlas-scale spatial reference maps such as HTAN |
Table 1. What spatial resolution adds to four questions a target discovery team asks routinely. Bulk data is not replaced; it is given spatial context.
Decoding the tumor microenvironment
Oncology is the application area where spatial biology has moved furthest into practical use, and the reason is structural rather than a matter of fashion. A tumor is not a uniform mass of cancer cells; it is an ecosystem of tumor cells, immune cells, stromal cells and blood vessels, and the interactions between those populations, not any one population’s abundance alone, determine how the tumor behaves and how it responds to treatment.
Multiplexed imaging technologies, including multiplexed ion beam imaging, imaging mass cytometry and cyclic immunofluorescence platforms, can now detect dozens to more than 100 protein targets simultaneously in a single tissue section. That capability is what makes characterizing the tumor immune microenvironment in full, rather than one or two markers at a time, a routine analysis instead of a specialized one-off study.
A 2025 study in Nature Genetics illustrates what that capability delivers concretely. Researchers applied spatial multi-omics profiling to 234 patients with advanced non-small cell lung cancer treated with programmed death 1-based immunotherapy across three independent cohorts. Using spatial proteomics, they identified a resistance cell-type signature, including proliferating tumor cells, granulocytes and vessels, associated with a hazard ratio of 3.8, and a response signature, including M1 and M2 macrophages and CD4 T cells, associated with a hazard ratio of 0.4. They then generated a cell-to-gene resistance signature from spatial transcriptomics that remained predictive of poor outcomes across all three cohorts, with hazard ratios of 5.3, 2.2 and 1.7 respectively.
Why the multi-cohort result mattersA signature discovered in one patient cohort and never checked against another is a hypothesis. A signature that keeps predicting outcome across three independent cohorts, even with the effect size varying as it does here, is evidence a biomarker development program can actually build on. That distinction, reproducibility across independent cohorts rather than significance within a single one, is what separates a promising finding from a candidate biomarker. |
The commercial and translational interest this generates is substantial, and our earlier coverage in The promise of spatial proteomics traces how mapping proteins within their native tissue environment is letting researchers connect molecular function to cellular architecture across a widening range of diseases beyond oncology alone.
Spatial biomarkers and companion diagnostics: the next frontier
This is the area where the field’s promise and its current regulatory reality are furthest apart, and stating that gap precisely is more useful to a biomarker program than glossing over it.
Immune checkpoint inhibitors have reshaped cancer therapeutics, but the population that benefits remains small in many tumor types, and average response rates continue to sit in the 20% to 30% range. Only a handful of companion diagnostic modalities are currently approved to help select patients for this therapy: immunohistochemistry for programmed death-ligand 1, immunohistochemistry or polymerase chain reaction testing for mismatch repair status, and polymerase chain reaction testing for microsatellite instability. Each has real, documented, limited predictive value, which is precisely the gap driving interest in a new biomarker class.
Spatial phenotypic signatures, defined by the measurement of cell densities and interactions between tumor and immune cells rather than expression level alone, are that new class. They are not, at this point, themselves an approved companion diagnostic category; they are the leading candidate for closing a documented predictive gap in the diagnostics that are approved. That distinction matters for anyone building a biomarker program on this technology: the platform is mature enough for research use and increasingly for clinical trial stratification, but the regulatory validation path for a spatial signature as a standalone companion diagnostic is still being built.
Diagnostic dimension | Traditional single-marker IHC | Spatial phenotypic signature |
What it measures | Presence and intensity of one protein | Density and spatial interaction of multiple cell populations |
Regulatory status | Multiple approved companion diagnostics exist | An emerging class, not yet an approved standalone modality |
Predictive power for immunotherapy response | Documented as limited on its own | Under active investigation specifically to close that gap |
Typical current use | Patient selection in the approved indication | Clinical trial stratification and biomarker discovery |
Table 2. Where spatial phenotypic signatures sit relative to currently approved single-marker diagnostics. The comparison is deliberately precise about regulatory status rather than implied.
One proof-of-concept approach worth understanding illustrates where this is headed: combining spatially resolved protein information from a digital spatial profiling platform with bulk transcriptomic data from the same patient cohort produced predictive models that exceeded the value of either data type alone, demonstrated in a cohort of immunotherapy-treated melanoma patients. That pattern, combining a spatial readout with a complementary molecular layer rather than relying on either in isolation, is likely to define how spatial biomarkers eventually earn a formal companion diagnostic role.
The path to the clinic
Moving a spatial biology finding from a discovery-stage dataset into a clinical workflow is a distinct challenge from generating the finding in the first place, and it is where the field’s next hard problems sit.
Three requirements dominate that transition.
- Standardized, validated assays. A discovery-stage multiplexed imaging panel optimized in one lab has to be re-established as a validated clinical assay, with defined analytical sensitivity, specificity and reproducibility, before it can support a treatment decision.
- Analysis pipelines that scale. The computational bottleneck in multiplexed tissue imaging is well recognized: the technology can generate massive datasets faster than most laboratories can analyze and interpret them, which is exactly why atlas-scale consortium tool development, of the kind behind CyLinter and CalicoST, matters beyond the academic groups that built them.
- Workflows that fit existing pathology practice. A spatial assay that requires a fundamentally different specimen-handling workflow than a pathology lab already uses faces an adoption barrier independent of its scientific merit.
Practical operational guidance on exactly this translation challenge, from platform selection through data throughput to workflow integration, is covered in our earlier piece, Optimizing spatial proteomics workflows for discovery and translation, which notes that given the complexity of spatial biology analysis, robust computational resources, software and expertise are vital for managing and analyzing the resulting data effectively.
Digital pathology is the discipline where this translation is furthest along, since it already sits at the intersection of tissue-based diagnostics and computational analysis, and it is the natural entry point for spatial methods moving toward routine clinical use. That specific translation path, and what is required to move a spatial assay from a research tool into a validated clinical one, is the subject of its own dedicated guide within this cluster.
Where spatial biology is headed in drug discovery
Three trends are shaping the next phase of adoption, and each has direct consequences for a drug discovery program deciding when and how to invest.
- Multi-omic integration is becoming the default rather than the exception. Combining spatial transcriptomics, spatial proteomics and increasingly spatial metabolomics on the same tissue, rather than choosing one modality, is where the strongest biomarker signatures are emerging, consistent with the multi-cohort lung cancer result described above.
- Artificial intelligence (AI) is being built into the analysis layer, not bolted onto it afterward. Given the scale of multiplexed imaging datasets, AI-driven approaches to segmentation, cell typing and pattern discovery are increasingly treated as core infrastructure rather than an optional add-on, a shift that mirrors how the computational bottleneck described above is actually being addressed in practice.
- The application space is expanding beyond oncology. While tumor biology remains the largest current use, the same tissue-architecture logic applies to neurodegeneration, autoimmune disease, and other conditions where cell-cell interaction in a specific tissue context drives pathology, a shift already visible in the range of disease areas this cluster’s emerging therapeutic areas guide will cover.
For a drug discovery organization weighing when to adopt spatial methods more deeply, the honest answer from where the field stands today is that the discovery and mechanism-understanding value is already well established, oncology biomarker development is maturing quickly but has not yet produced an approved spatial companion diagnostic, and the translational and clinical workflow questions are the genuine frontier. Each of those three statements points to a different part of this cluster, and each is covered in full in its own dedicated guide.
This article was produced under Drug Discovery News’s editorial policies.
