Applying single-cell spatial transcriptomics to build a disease map means treating single-cell and spatial data as two answers to two different questions rather than as competing technologies. One tells a discovery program which cell types and states exist in a diseased tissue. The other tells it where those cells sit, and next to whom, which is frequently the detail that turns a catalog of cell types into an actual mechanistic hypothesis.
Key takeaways
|
The single-cell foundation
Single-cell RNA sequencing remains the starting point for most disease-mapping efforts, because it does something no earlier bulk method could: it resolves gene expression at the level of an individual cell, revealing cell types, cell states and transitions between them that a tissue-level average would blend together entirely.
Our earlier coverage of the field, in Illuminating the transcriptomic landscape in single cancer cells at scale, traces how far that resolution has come as sequencing throughput and cost have improved. What single-cell data cannot do, by the nature of how it is generated, is tell a researcher where in the tissue a given cell was positioned. Dissociating tissue into individual cells for sequencing is the step that makes single-cell resolution possible, and it is also the step that discards spatial position entirely, a trade-off covered in depth in Spatial biology for target discovery: Mapping disease at tissue resolution.
That trade-off is why single-cell data functions best as a foundation rather than a finished picture. It establishes the cast of cellular characters present in a disease; it does not establish the scene they occupy.
Adding spatial context
Spatial transcriptomics and related spatial omics methods complete the picture single-cell data leaves unfinished, and the cleanest statement of the relationship is a direct one from the methods literature.
The question spatial data was built to answerSpatial transcriptomics answers a question that standard single-cell RNA sequencing cannot: where, within an intact tissue section, a given transcript originates. That is a precise way to state the complementary relationship between the two, since it avoids the common but misleading framing that spatial methods are simply an upgraded or higher-resolution version of single-cell sequencing. The two are not competing on the same axis. Single-cell sequencing generally reaches deeper into the transcriptome or profiles more cells per dollar than most spatial methods currently manage; spatial methods answer the positional question single-cell data structurally cannot address at all, regardless of sequencing depth or cost. A discovery program does not choose one over the other so much as decide which question it needs answered first. |
Platform mechanics for how that spatial context is actually captured, including the distinction between sequencing-based platforms that bin transcripts into spots and imaging-based platforms that resolve individual molecules at subcellular resolution, are covered in detail by our colleagues at Technology Networks in Spatial transcriptomics: Methods and platforms guide, which is the right next stop for a reader who needs to choose a specific platform rather than understand the strategic case for combining data types.
Building disease reference maps
The most ambitious application of combined single-cell and spatial data is the reference atlas: a shared, community-scale map of a disease built from many patients, many disease stages and both data types together, intended to be queried by other researchers rather than analyzed once and set aside.
The Human Tumor Atlas Network is the clearest example of this approach applied to cancer specifically. Its founding framework describes generating single-cell, multiparametric, longitudinal atlases and integrating them with clinical outcomes as a route to identifying novel predictive biomarkers and features, as well as therapeutically relevant cell types, cell states and cellular interactions across disease transitions. The stated translational promise spans basic understanding of disease mechanisms, diagnosis, prognosis, treatment monitoring, drug development, biomarker discovery and patient stratification, explicitly naming drug development as one of the atlas’s intended uses rather than an incidental byproduct.
A recent example shows what that translational promise looks like in a specific, current study. Researchers combined single-nucleus RNA and ATAC sequencing with spatial omics on matched pre- and post-chemotherapy samples from 22 patients with high-risk neuroblastoma, characterizing therapy-related changes in both cell composition and the tumor microenvironment directly. That design, paired samples from before and after treatment, combined with spatial data rather than sequencing alone, is what lets a research team distinguish a genuine treatment effect on the tissue from the kind of patient-to-patient variability a single-timepoint, single-modality study cannot rule out.
Reference map component | What single-cell data contributes | What spatial data adds |
Cell type catalog | Comprehensive identification of cell types and states present across many patient samples | Confirmation that a given cell type occupies a consistent tissue location across patients, not just a consistent expression profile |
Disease transitions | Trajectory inference showing how cell states change as disease progresses, based on expression similarity | Direct evidence of where those transitions occur physically, and which neighboring structures are present at each stage |
Treatment response | Before-and-after comparison of cell type proportions and expression, as in the neuroblastoma example above | Whether treatment-associated changes are localized to specific tissue regions or occur uniformly throughout |
Community reuse | A queryable reference for cell type identity that other researchers can compare their own data against | A queryable reference for spatial context that grounds cell type identity in tissue architecture, not expression alone |
Table 1. What each data type contributes to a shared disease reference map. The value of combining them is cumulative rather than either type standing in for the other.
Applications in discovery
Three uses of a combined single-cell and spatial disease map recur across current drug discovery programs.
- Target discovery grounded in both identity and position. A candidate target’s cell-type specificity, established by single-cell data, gains an additional layer of evidence when spatial data confirms it occupies a consistent, disease-relevant location across patients rather than appearing in scattered, inconsistent positions.
- Mechanism studies that track a process across a tissue directly. Rather than inferring a spatial process, such as an immune cell’s migration toward a tumor, from expression data alone, combined data can show that process directly as a physical gradient across the tissue.
- Treatment response characterization using matched, longitudinal samples. The neuroblastoma chemotherapy example above illustrates this directly: understanding how a specific treatment reshapes both cell composition and tissue architecture, in the same patients, before and after exposure.
Each of these applications depends on the atlas or reference map being genuinely queryable rather than a one-time analysis, since the value of a shared community resource compounds as more researchers compare their own data against it and as more disease stages and patient populations are added over time.
Data and integration challenges
Generating both data types is only the first obstacle. Reconciling them computationally is frequently the actual bottleneck in building a usable disease map, and it is worth naming the specific difficulties rather than treating integration as a solved, routine step.
- Resolution mismatch between data types. Single-cell data reports one profile per dissociated cell; many spatial platforms report one profile per spot or region containing several cells, and reconciling those units requires an explicit computational strategy rather than a simple merge.
- Batch effects across patients and platforms. A reference atlas assembled from many research groups, instruments and protocols accumulates technical variation that has to be corrected before biological signal can be trusted across the combined dataset.
- Annotation consistency across contributing studies. A cell type labeled one way in one contributing dataset and a different way in another has to be reconciled before the combined atlas is queryable as a single coherent resource rather than a collection of incompatible datasets.
- Long-term maintenance and versioning. A living, growing reference atlas needs an explicit plan for incorporating new data and tracking changes over time, a governance question as much as a technical one.
For where building disease maps sits within the broader tissue-context argument for target discovery, see Spatial biology for target discovery: Mapping disease at tissue resolution, and for where this fits within the full spatial biology pipeline, Spatial biology in drug discovery: From target discovery to translational medicine.
This article was produced under Drug Discovery News’s editorial policies.
