New brain mapping method profiles 40 million cells to reveal genetic differences

Researchers have introduced a tissue-mapping method called ATLAS that can chart cell types across entire mouse brains at a scale that was previously impractical, and they used it to profile 40 million cells in search of transcriptional differences linked to genetics and sex. The work, published in Nature Methods, addresses a long-standing bottleneck: comparing organ structure across genetic backgrounds requires spatial, single-cell, atlas-scale data collected in replicates, a demand that conventional imaging-based transcriptomics struggles to meet.

ATLAS, short for atlas-scale transcriptome localization using aggregate signatures, takes a different route from methods that measure thousands of genes directly in tissue. The approach learns transcriptional signatures from single-cell RNA sequencing data, encodes those signatures in situ with tens of thousands of oligonucleotide probes, and then decodes the measurements computationally to infer cell types and imputed transcriptomes. In other words, the method leverages existing atlas data and in silico decoding to generate imputed cell maps rather than reading out every gene in place.

The team validated the technique against direct MERFISH measurements of marker genes and through quantitative comparisons with four other technologies, giving the imputed maps an empirical benchmark rather than relying on simulations alone. The validation step matters because imputed transcriptomes are statistical reconstructions; showing that they agree with direct measurements of known markers is what makes downstream biological comparisons credible.

With the method in hand, the researchers profiled more than 400 coronal sections from the central brains of five male and five female C57BL/6J mice, an inbred strain widely used as a laboratory reference, and five male BTBR T+tf/J mice, a strain used as a model of idiopathic autism. The resulting dataset covered the central brain in three dimensions, allowing the team to infer cell-type distributions across the B6-female, B6-male and BTBR-male conditions and to compare them systematically.

The comparative analysis surfaced substantial biological signal. The researchers reported more than 40 significant differences in cell-type distributions and identified 16 regional composition changes across the male–female and B6–BTBR comparisons. These findings, visualized as spatial differences in cell-type distribution and in regional composition, indicate that both genetic background and sex leave detectable fingerprints in the cellular makeup of brain regions.

The BTBR comparison is particularly notable for neuroscience. Because BTBR T+tf/J mice display autism-relevant behaviors without a single known causal mutation, the strain is a common tool for studying idiopathic neurodevelopmental disease. ATLAS now provides a way to examine how the cellular composition of the BTBR brain differs from the C57BL/6J reference across whole regions, rather than in a handful of manually chosen sections, which could help connect behavioral phenotypes to specific spatial changes in brain cell populations.

Technically, the method’s efficiency comes from its encoding strategy. Instead of building a probe panel that resolves every transcript individually, ATLAS encodes aggregate transcriptional signatures learned from reference single-cell RNA sequencing atlases. Tens of thousands of oligonucleotide probes capture enough information in situ for the decoding algorithm to assign cell types and reconstruct imputed transcriptomes. This design keeps the experimental burden manageable while retaining atlas-scale coverage, which is what makes replicate-based comparative studies across multiple animals feasible.

The authors position ATLAS as a tool for systematic comparative studies that facilitate organ-level structure–function analysis of disease models. Because genetic variation can alter organ structure and function, having a scalable way to generate spatial, single-cell maps in replicates removes a key obstacle to asking how a mutation, a strain difference or a disease model reshapes an organ’s cellular architecture. The 40-million-cell dataset spanning five animals per condition demonstrates that such comparisons can now be run at whole-organ scale.

Limitations remain inherent to the design. The decoded transcriptomes are imputed from reference atlases, so their fidelity depends on the quality and relevance of the underlying single-cell RNA sequencing data, and the validation against MERFISH covers marker genes rather than the full transcriptome. The published study also focused on mouse central brains in specific strain and sex comparisons, so extending the approach to other organs, species or disease contexts will require appropriate reference atlases and further benchmarking.

Even with those caveats, the study marks a shift in what spatial transcriptomics can deliver. By combining in situ encoding with computational decoding, ATLAS turns atlas-scale comparative neuroanatomy from a bespoke, resource-intensive effort into a repeatable workflow, and the more than 40 cell-type differences and 16 regional composition changes it revealed in this first application suggest that such workflows will keep finding biology that section-by-section studies have missed.

Subject of Research: A scalable spatial transcriptomics method that maps cell types across mouse brains to compare genetic backgrounds.

Article Title: Spatial single-cell mapping of transcriptional differences across genetic backgrounds in mouse brains

Article References: Hemminger, Z., Tam, G., De Ocampo, H., Wang, A., Underwood, T., Xie, F., Zhao, Q., Song, D., Li, J. J., Dong, H., & Wollman, R. (2026). Spatial single-cell mapping of transcriptional differences across genetic backgrounds in mouse brains. Nature Methods. https://doi.org/10.1038/s41592-026-03289-2

Image Credits: AI Generated

DOI: 10.1038/s41592-026-03289-2

Keywords: spatial transcriptomics, ATLAS, MERFISH, single-cell RNA sequencing, mouse brain, cell-type distribution, BTBR autism model, C57BL/6J, brain atlas, in situ decoding, organ-level analysis, Spatial