The case for drug target identification in tissue context, rather than in cells pulled apart from it, now rests on more than intuition. A retrospective analysis of known drug targets across dozens of diseases found that a target’s expression pattern in intact, disease-relevant tissue is one of the more reliable predictors available of whether that target will actually survive clinical development, a finding with direct consequences for how early-stage target discovery programs should be built.
Key takeaways
|
Why tissue context changes target biology
Single-cell RNA sequencing transformed target discovery by revealing which specific cell types express a gene of interest, a major advance over bulk tissue measurements that could only report an average across an entire mixed sample. But the standard single-cell workflow requires dissociating tissue into a suspension of individual cells before sequencing, and that step destroys exactly the information that determines whether a cell type’s expression pattern is therapeutically meaningful: its position relative to everything around it.
Two consequences follow directly from that loss, and both change how a target discovery team should read single-cell data.
- Proximity is invisible after dissociation. A receptor expressed on a tumor cell adjacent to exhausted immune cells represents a different therapeutic opportunity than the identical receptor on a tumor cell with no immune contact at all, but dissociated single-cell data cannot distinguish the two, because the positional relationship no longer exists once the cells are separated.
- Rare, spatially restricted populations can be underrepresented or lost. Cell types that depend on a specific microenvironment to survive dissociation, or that exist only within a defined spatial niche, are more likely to be lost or diluted during tissue processing than in an assay that measures gene expression directly within the intact tissue.
Spatial transcriptomics addresses this directly by preserving locational information during measurement itself. As one recent methods review puts it, spatial labeling assigns unique locational barcodes to transcripts within intact tissues, which is the specific technical mechanism that keeps position attached to expression data rather than discarding it. The same review notes the trade-off plainly: spatial transcriptomics provides valuable spatial information on gene expression in tissue sections but can lack the single-cell resolution required to distinguish individual cell types, while single-cell RNA sequencing provides cellular-level data but loses locational information due to tissue dissociation. Neither approach alone is complete, which is why combining them, covered later in this hub, has become standard practice rather than a specialized technique.
Finding targets in spatial data
This is where the evidence moves from plausible to quantified. A retrospective analysis of known drug target genes examined whether successful targets, meaning those that reached clinical development and progressed through trials, were associated with two specific expression patterns detectable in single-cell and spatial data: cell-type-specific expression in a disease-relevant tissue, and cell-type-specific over-expression in disease patients compared with healthy controls.
A specific, quantified answer to whether context predicts successAnalyzing scRNA-seq data across 30 diseases and 13 tissues, the study found that both cell-type specificity and disease-cell specificity significantly increase the odds of clinical success for a gene-disease pairing. Cell-type specificity carried an odds ratio of 2.47, and disease-cell specificity an odds ratio of 2.34, both associations statistically robust across the dataset. Combined, the two forms of evidence were estimated to approximately triple the chances of a target reaching phase III. The same analysis found that this tissue-context evidence identifies a target space that is larger than, and complementary to, the space identified by direct genetic evidence alone, and that it is more likely to prioritize therapeutically tractable target classes such as membrane-bound proteins. In practical terms, a target discovery program that checks candidates against tissue-context expression data is working from a broader and more actionable pool of evidence than genetics alone provides. |
That finding reframes what a spatial or single-cell dataset is actually for at the discovery stage. It is not only a tool for generating hypotheses about which genes might matter; the expression pattern itself, specifically whether it is concentrated in a disease-relevant cell type and disease state, functions as a predictor a team can weigh alongside genetic evidence when prioritizing which candidates to advance.
Spatial validation of targets
Validation is a different question from discovery. Discovery asks which genes might matter; validation asks whether a specific candidate, once identified, actually behaves the way a therapeutic hypothesis requires once it is examined in its native tissue context.
Three checks define spatial validation, and each addresses a way a target can look promising in isolation and fail in context.
- Confirm cell-type specificity in situ. Does the target remain concentrated in the expected cell type when measured directly in tissue, or does apparent specificity in dissociated data turn out to reflect a processing artifact rather than a genuine biological pattern?
- Confirm spatial proximity to the relevant structure. If the therapeutic hypothesis depends on the target cell interacting with a specific neighbor, such as an immune cell, a vessel, or a specific tissue compartment, spatial data can confirm or rule out that proximity directly rather than assuming it.
- Confirm the pattern reproduces across patients. A spatial signature identified in one patient sample is a hypothesis. Whether it recurs across a patient cohort, rather than reflecting one unusual sample, is what separates a validated target signature from a promising but unconfirmed observation.
That third check connects directly to a broader methodological caution worth carrying into any target validation program. A scoping review of spatial transcriptomics in cancer research examined 41 published studies and found that nearly half lacked comprehensive data processing protocols, a gap the reviewers noted hinders reproducibility. The practical lesson for a target validation program is to document processing choices as rigorously as the biological findings themselves, since an unreproducible pipeline undermines confidence in a validated target just as much as an unreproducible biological signal does.
Disease heterogeneity and targets
Most diseases are not biologically uniform across patients, or even across regions of the same affected tissue, and a single-target strategy built on an assumption of uniformity is vulnerable to exactly the heterogeneity spatial data is built to reveal.
A 2024 study in idiopathic pulmonary fibrosis, a progressive lung disease with only two approved antifibrotic therapies and a persistent unmet need, illustrates how directly this translates into new target hypotheses. Researchers profiled IPF and control lung tissue using spatial transcriptomics and integrated the data with an existing single-cell RNA sequencing atlas of the disease. They identified three distinct disease-associated niches, each with its own cellular composition and tissue location: a fibrotic niche made up of myofibroblasts and aberrant basaloid cells located around airways, an adjacent airway macrophage niche in the airway lumen containing a specific macrophage subtype, and a separate immune niche characterized by distinct lymphoid cell clusters surrounded by remodeled blood vessels.
The researchers state their conclusion directly: this spatial characterization of IPF niches will facilitate the identification of drug targets that disrupt disease-driving niches and will aid development of disease-relevant laboratory models. That is target discovery working exactly as the tissue-context argument predicts, in a disease area where dissociated single-cell data had already been extensively used but had not, on its own, resolved which cellular interactions were actually driving pathology.
Heterogeneity source | Risk to a single-target strategy | What spatial data adds |
Regional variation within one tissue | A target validated in one region may be absent or irrelevant elsewhere in the same organ | Maps where a target is actually expressed across the full tissue, not one sampled region |
Disease-associated niches | A therapy may address one pathological process while an adjacent, distinct niche continues driving disease | Identifies multiple disease-driving niches separately, as in the IPF fibrotic, macrophage and immune niches |
Patient-to-patient variability | A signature from one patient sample may not generalize to the treatment population | Supports checking whether a spatial pattern reproduces across a patient cohort before committing to it |
Cell-cell interaction dependency | A target’s relevance may depend on a neighboring cell type that bulk or dissociated data cannot reveal | Directly measures which cell types are physically adjacent and can be analyzed for their communication |
Table 1. Four sources of disease heterogeneity that complicate a single-target strategy, and what spatial resolution adds to each. The IPF niche example demonstrates the second row concretely.
From spatial signal to program
A statistically significant spatial pattern is not yet a drug discovery program, and the gap between the two is where many promising findings stall. Four steps generally separate an interesting spatial observation from a candidate a team can commit resources to advancing.
- Establish reproducibility across patients. As covered above, a pattern confirmed in a single sample is preliminary until it recurs across an independent patient cohort.
- Cross-reference against tissue-context clinical success evidence. Checking a candidate’s cell-type and disease-cell specificity against the kind of evidence described earlier is a low-cost way to weigh a candidate before committing to expensive validation work.
- Confirm the mechanism implied by proximity. If the spatial signal suggests a specific cell-cell interaction drives pathology, that interaction generally needs functional confirmation, not only positional correlation, before it supports a therapeutic hypothesis.
- Document the analysis pipeline alongside the biological finding. Given how commonly published spatial studies omit full processing detail, a program-ready target package should include the computational methods used to reach a conclusion, not only the conclusion itself.
Multi-omic integration, combining spatial transcriptomics with single-cell data, proteomics or both on the same or matched tissue, is increasingly how the strongest of these signals are generated rather than relying on a single spatial modality alone. That integration, and what the combined data actually shows once the two are brought together, is treated in full in Spatial Multi-Omics for Target Discovery: What the Data Actually Shows. The mechanics of moving from tissue in hand to usable target-discovery data, including how targets are actually located and prioritized from spatial datasets, are covered in Finding Drug Targets in the Tissue Context With Spatial Biology. Disease heterogeneity beyond the IPF example above, including its particular weight in oncology, is developed further in Understanding Disease Heterogeneity With Spatial Biology. And the practical question of when to choose single-cell data, spatial data or both, and how to build an integrated disease map from either, is addressed in From Single-Cell to Spatial: Building Disease Maps for Drug Discovery.
For where target discovery sits within the wider spatial biology pipeline, from mechanism 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.
