Traditional gene-expression analysis often treats cells as isolated collections of molecular measurements. Spatial technologies such as Xenium add a crucial layer of information by recording the physical positions of RNA molecules within tissue sections. Yet this richer data creates a computational challenge: a cell may contain thousands of transcripts distributed across different compartments, while its behavior can also depend on nearby immune, epithelial, stromal or malignant cells. SpatialFormer was designed to integrate these signals instead of analyzing them separately. Its objective is to produce a compact numerical representation of each cell that reflects both its molecular identity and its location within a tissue niche.

The model combines two complementary forms of machine learning. Convolutional networks are well suited to recognizing local patterns, such as the arrangement of transcripts within a cell or the immediate organization of neighboring cells. Transformers, by contrast, use attention mechanisms to evaluate relationships across a broader context. In SpatialFormer, these capabilities are used to capture information at both subcellular and multicellular scales. The result is intended to preserve fine-grained spatial gene-expression patterns while also representing the larger cellular environment in which those patterns acquire biological meaning.

To train the system, the researchers used a pairwise strategy based on relationships between cells. Rather than learning from a single cell in isolation, the model was exposed to cell pairs and their associated spatial and expression information. This approach allows it to distinguish cellular features that are intrinsic to a cell from signals that emerge through proximity or shared niche context. During pretraining, SpatialFormer processed approximately 700 million cell pairs drawn from 17 million spatially resolved single cells across 71 Xenium slides. The scale of this dataset is central to the project: exposure to diverse tissues and cellular arrangements can help the model learn reusable biological patterns rather than memorizing one experimental sample.

The training data came from Xenium spatial transcriptomics, a technology that detects selected RNA molecules while retaining their coordinates inside tissue. In practical terms, this means that researchers can ask whether a gene is concentrated near the nucleus, enriched toward a cell boundary, or distributed in a pattern associated with a particular cellular state. They can also examine whether cells expressing complementary or competing molecular programs are positioned next to one another. SpatialFormer uses these measurements as more than a list of gene counts, treating them as structured spatial signals that can be translated into representations for downstream analysis.

The authors report that these learned representations supported several important tasks in single-cell and spatial biology. In batch correction, the model helped reduce technical differences between experiments while retaining biological variation, an essential step when datasets are generated on different slides, in different laboratories or under different conditions. In cell-type annotation, the representation provided molecular and contextual information that can assist in assigning identities to cells. The framework was also used for co-localization detection, identifying cell populations that repeatedly occupy the same tissue regions or appear to participate in shared microenvironments.

Co-localization is particularly significant because physical proximity can provide clues about communication, competition or coordinated function. A neighboring relationship alone does not prove that two cell types directly interact, but it can identify tissue regions where signaling is biologically plausible. By incorporating gene expression and spatial arrangement together, SpatialFormer may help researchers move beyond maps that simply label cells toward models that explain why particular cell types gather in specific locations. This could be valuable in tissues where disease progression depends on complex local interactions rather than on the behavior of one cell population alone.

The study also used perturbation analysis to investigate which molecular signals may be especially important in disease-associated cellular relationships. In pulmonary fibrosis, the analysis identified gene pairs linked to immune cell–cell communication. In breast cancer, it highlighted signals associated with epithelial–myoepithelial co-localization and tumor transition states. These findings suggest that the model can be used to computationally test the importance of individual genes or gene pairs within a spatial network. Such perturbation analyses do not automatically establish clinical causation, but they can prioritize molecular interactions for laboratory experiments and potentially reveal mechanisms that would be difficult to detect from expression levels alone.

SpatialFormer’s broader promise lies in its attempt to create a common language for spatial biology. Researchers currently work with datasets that differ in tissue type, measurement technology, resolution and experimental design. A representation learned from millions of spatially resolved cells could make it easier to compare these datasets and transfer knowledge between studies. If successful across additional tissues and platforms, systems of this kind could support the discovery of disease-specific niches, improve the interpretation of tissue biopsies and help identify cellular interactions that influence treatment response. The framework does not replace biological validation, but it offers a high-dimensional map of cellular relationships that may guide the next generation of experiments in immunology, fibrosis and cancer research.

Subject of Research: Spatial representation learning for single-cell multimodal and multiscale gene expression, cellular niches and spatial cell–cell interactions.

Article Title: SpatialFormer: universal spatial representation learning from subcellular molecular to multicellular landscapes

Article References: Wang, J., Huang, Y. & Winther, O. SpatialFormer: universal spatial representation learning from subcellular molecular to multicellular landscapes. Nat Comput Sci (2026). https://doi.org/10.1038/s43588-026-01016-7

Image Credits: AI Generated

DOI: https://doi.org/10.1038/s43588-026-01016-7

Keywords: SpatialFormer, spatial transcriptomics, single-cell biology, gene expression, cellular niches, transformers, convolutional neural networks, Xenium, cell–cell communication, pulmonary fibrosis, breast cancer, tumor biology

sc-SPLASH represents a transformative shift in the analytical paradigm of scRNA-seq data. Instead of relying on prior knowledge encoded in existing genomic references, this method harnesses barcoded data in a manner that identifies novel transcriptomic features independent of alignment. This is especially crucial when analyzing species with incomplete or missing reference genomes, such as the sponge Spongilla and the tunicate Ciona, organisms that occupy pivotal evolutionary positions but have been historically genomically understudied. The ability to uncover previously unknown genomic elements and transcript variants in such organisms opens new avenues in evolutionary biology and functional genomics.

Central to the sc-SPLASH framework is its BKC submodule, a component engineered to optimize preprocessing of barcoded sequencing data. Preprocessing of such data—especially unique molecular identifiers (UMIs) which are critical for eliminating amplification bias and counting transcripts accurately—can be a computational bottleneck. Remarkably, BKC has been demonstrated to operate approximately 50 times faster than the commonly used UMI-tools pipeline, commonly considered an industry standard. This incredible improvement in speed and efficiency will enable researchers to process larger datasets more swiftly, accelerating discovery cycles and reducing computational resource demands.

The reference-free aspect of sc-SPLASH leverages sophisticated statistical models that detect transcriptomic variation at the barcode level before any alignment steps take place. This approach allows for an unbiased detection of features such as secreted repeat proteins, which can be obscured or entirely missed when relying solely on reference-guided methods. In their analyses, the authors discovered immune-like cells in both Spongilla and Ciona exhibiting expression of secreted repeat proteins that were notably absent from existing reference annotations. These findings suggest a rich layer of functional complexity previously masked by conventional analytical pipelines.

From a technical perspective, sc-SPLASH integrates advanced algorithms capable of handling the massive scale and complexity of barcode-laden single-cell datasets. The method efficiently resolves barcode errors and PCR duplicates, crucial for ensuring data integrity and reducing false positives. Moreover, it captures subtle transcriptomic features that could be lost in noise or mistaken for artifacts in other methods. This is achieved through a carefully designed statistics-first workflow that prioritizes authentic biological signal from the outset.

The implications of deploying sc-SPLASH extend beyond single-cell research into spatial transcriptomics, where gene expression is mapped within tissue architecture. By applying a reference-free analysis, spatial transcriptomic studies can unveil novel cell types, states, or spatially restricted transcript variants that defy current genomic annotations. This capability enhances our understanding of tissue complexity and cellular interactions, particularly in non-model organisms and novel experimental contexts.

Notably, the authors emphasize that sc-SPLASH empowers open-ended discovery. Without preconceptions imposed by incomplete or biased reference genomes, researchers are freer to uncover unanticipated biology. This is particularly relevant in evolutionary and environmental biology, where many species lack comprehensive genomic resources. By revealing new classes of proteins and transcript variants, even in well-studied organisms, this tool opens the door for novel hypotheses about cellular function and evolution.

The dramatic speed improvement from the BKC submodule means that large-scale studies involving hundreds of thousands to millions of cells—the scale at which many cutting-edge single-cell atlases operate—become feasible within reasonable timeframes and computing budgets. This gains paramount importance as scRNA-seq experiments increasingly push towards whole-organism or multi-organ datasets that generate vast volumes of barcoded reads.

In addition, sc-SPLASH’s focus on barcoded data acknowledges the modern realities of single-cell sequencing. Barcodes, including cell and molecule identifiers, are crucial for deconvoluting complex data but also introduce noise and error. The methodological sophistication of sc-SPLASH in preprocessing these barcodes ensures that downstream biological inference is not compromised, establishing a new benchmark for data quality and reliability in the field.

Perhaps most exciting, the application of sc-SPLASH to uncover immune-like cells expressing secreted repeat proteins in relatively unexplored metazoans like Spongilla and Ciona underscores how new computational tools can revive biological inquiry into basal animal lineages. These discoveries are not just esoteric; they have the potential to reshape our understanding of immune system evolution, the diversification of repeat protein functions, and possibly biotechnological applications where novel repeats may serve as scaffolds or bioactive molecules.

As single-cell sequencing technologies evolve and expand, tools like sc-SPLASH are essential to harness the full information content inherent in these high-dimensional datasets. By eliminating reliance on incomplete references, reducing computational costs, and focusing on intrinsic statistical properties of barcoded data, this method sets a new standard for exploratory transcriptomics. It invites a move away from the constraints of known gene catalogs towards a more holistic, unbiased exploration of cellular identity and function.

Furthermore, the integration of sc-SPLASH with existing computational pipelines will be straightforward for many laboratories. Its design emphasizes compatibility with raw barcoded reads and modularity in preprocessing, meaning that even well-established workflows can benefit from its optimized speed and statistical rigor. This accessibility will likely accelerate its adoption and catalyze discoveries across diverse biological fields.

The publication of this method at a time when single-cell and spatial transcriptomics are becoming foundational across basic, translational, and clinical research domains ensures its broad relevance. From decoding microbial communities to exploring human disease heterogeneity, reference-free discovery methods fill a critical gap by overcoming the limitations of incomplete or absent genomic references.

In conclusion, sc-SPLASH represents a leap forward in the analytical toolkit available to the life sciences community. By enabling reference-free, statistics-first interrogation of barcoded single-cell and spatial transcriptomic data, it opens new vistas for discovery in both model and non-model organisms alike. The ability to detect hidden transcriptomic complexity with unprecedented speed and accuracy heralds a new age of functional genomics that prizes unbiased exploration as much as hypothesis-driven inquiry.

This innovative approach not only enhances our capacity to characterize novel cell types and molecular players but also challenges the community to reconsider the dominance of reference-based paradigms in the era of big data biology. The future of single-cell research will undoubtedly be shaped by such advances that allow the data to speak for itself, unshackled from the confines of prior knowledge.

Subject of Research: Reference-free, barcoded single-cell RNA sequencing and spatial transcriptomics data analysis.

Article Title: Reference-free discovery with barcoded single-cell sequencing.

Article References:
Dehghannasiri, R., Kokot, M., Starr, A.L. et al. Reference-free discovery with barcoded single-cell sequencing. Nat Biotechnol (2026). https://doi.org/10.1038/s41587-026-03084-6

Image Credits: AI Generated

DOI: https://doi.org/10.1038/s41587-026-03084-6