The core insight behind VIMA is that a small patch of tissue is itself an image, and image-analysis machinery is exceptionally good at summarizing images. The method first rasterizes every sample into 10-micrometer pixels, converts the raw markers into a smaller set of ‘meta-markers’ using smoothing and principal component analysis, and removes batch artifacts with the Harmony algorithm. A sliding window then cuts each sample into overlapping 40-by-40-pixel squares, each spanning tens to hundreds of cells. Crucially, rather than segmenting individual cells or assigning them to predefined types, VIMA treats each square as an indivisible unit of spatial biology and asks a neural network to compress it into a 100-dimensional numerical fingerprint — a compact code in which biologically similar patches land close together.

Three technical choices distinguish VIMA from recent efforts in segmentation-free tissue representation. First, the autoencoders are conditional: the encoder and decoder receive a learned embedding of the sample identity, so the variational penalty pushes the network to strip out sample-specific quirks such as staining artifacts while the decoder supplies them back during reconstruction. This forces the fingerprints to reflect biology rather than batch effects, a property the authors quantified using a perplexity-based integration score across samples. Second, because independent training runs of neural networks yield meaningfully different representations, VIMA trains an ensemble of ten identical ResNet18-style conditional variational autoencoders with different initial weights, capturing complementary views of the same dataset. Third, instead of clustering patches into hard, nonoverlapping groups, VIMA defines thousands of small, overlapping ‘microniches’ anchored at each patch, with membership weighted by the probability that a random walk through a nearest-neighbor graph reaches the anchor.

The statistical machinery then operates on a tensor of samples by microniches by autoencoders, called the microniche abundance tensor. VIMA computes the correlation between each microniche’s abundance and case-control status, meta-analyzes the ten correlations for each patch using an adaptive weighted average, and estimates empirical false discovery rates by permuting sample labels while preserving donor structure. The output is threefold: a global P value for any spatial difference between cases and controls, a set of individual tissue patches that drive the signal at a chosen FDR threshold, and directional effect sizes for every patch. Simulations spiked into real MERFISH data confirmed that the P values are properly calibrated and that the method retains power and spatial accuracy even at the modest sample sizes typical of research cohorts, with runtimes of roughly two to eight hours on a single GPU.

To prove the method works on real disease, the team applied VIMA to three datasets spanning different technologies and clinical questions. The first was a seven-marker immunofluorescence dataset of synovial biopsies from 22 rheumatoid arthritis patients, collected by the Accelerating Medicines Partnership consortium. Even though a much larger single-cell RNA sequencing study had previously grouped these patients into fibroblast-rich ‘stromal’ and immune-rich subtypes using roughly 20,000 genes, VIMA recovered the stromal-versus-immune separation from just seven markers, achieving a global P value of 2.0 × 10⁻⁴ with more than 5,000 significant patches. The stromal-associated patches were enriched for CLIC5, a lining fibroblast marker, while immune-associated patches carried CD3, CD68 and DAPI signal — exactly the biology one would expect.

The arthritis analysis also revealed something that sample-wide cell-type abundances cannot capture: immune-labeled samples often contained spatially segregated stromal-like and immune-like regions within the same biopsy. This challenges the notion that each patient carries a single disease subtype and suggests that precision-medicine approaches in rheumatoid arthritis may need to account for geography inside the joint. Subclustering of the associated patches further showed that the myeloid-predominant subtype is characterized by vascularization, macrophage–fibroblast interactions and discrete T cell aggregates, whereas the T cell–myeloid subtype features an expanded lining layer and a distinct activated T cell infiltrate — differences invisible to nonspatial profiling.

The second application examined ulcerative colitis using a 52-marker CODEX dataset of colonic biopsies from 34 patients. VIMA cleanly separated diseased from healthy tissue and, critically, was the only benchmarked method able to detect UC-associated patches rather than just healthy-associated ones — a prerequisite for studying treatment effects within diseased tissue. When the team restricted the analysis to UC-associated patches and compared patients with prior or chronic TNF inhibitor exposure against those with none or recent initiation, VIMA found a robust association. Untreated tissue showed B cell aggregates surrounded by T cells, the hallmark of lymphoid aggregates, while chronically treated tissue displayed a disorganized mesenchymal pattern with scattered lymphocytes and elevated CD54, CD90, CD34 and MMP12. The fraction of treatment-associated patches stratified patients along a dose-response-like gradient of TNF exposure, providing the strongest evidence yet that TNF inhibition remodels colonic architecture by disrupting lymphoid aggregation.

The third and perhaps most striking application targeted dementia. The team analyzed a 140-gene MERFISH dataset containing 516 million transcripts from 75 postmortem medial temporal gyrus samples. VIMA detected a significant association with dementia status despite the small cohort — something neither seven state-of-the-art spatial methods nor a nonspatial cell-type-proportion analysis could achieve, indicating that spatial information effectively reduces the sample size needed to find signal. Control-associated patches colocalized with cortical layers 2 and 3, consistent with known thinning of these layers in Alzheimer’s disease and validating the approach against established histology. But VIMA also surfaced something genuinely new: dementia-associated patches concentrated in layer 6b, the deepest cortical layer, were enriched for oligodendrocytes by a factor of 1.33 and depleted of L6 IT Car3 neurons. Because oligodendrocytes have recently been shown to produce amyloid-β in the cortex, and myelin dysfunction is implicated in amyloid deposition, the finding raises the possibility that the deleterious versus protective roles of these cells depend on spatial context — with amyloidogenic activity potentially localized near the white matter boundary.

Ablation studies confirmed that every ingredient matters. Removing the conditional design, the ResNet architecture, or the microniche framework in favor of clustering progressively degraded performance across all four phenotypes, with the subtler treatment and dementia signals being especially dependent on strong sample integration. The authors are careful to note the method’s limitations: it requires a GPU, patch size choices shape what signals are visible, and interpretation still demands downstream biological analysis. But by fusing image-inspired deep learning with permutation-based hypothesis testing, VIMA offers something spatial biology has lacked — a statistically principled way to ask where, precisely, disease lives in tissue, without forcing cells and niches into boxes that may not fit the biology. As spatial atlases multiply, tools of this kind may become the standard lens through which researchers translate tissue geography into clinical insight.

Subject of Research: Deep-learning-based case-control analysis of spatial molecular tissue data

Article Title: Accurate and well-powered case–control analysis of spatial molecular data

Article References: Reshef, Y. A., Sood, L., Curtis, M., Rumker, L., Stein, D. J., Palshikar, M. G., Nayar, S., Filer, A., Jonsson, A. H., Korsunsky, I., & Raychaudhuri, S. (2026). Accurate and well-powered case–control analysis of spatial molecular data. Nature Methods. https://doi.org/10.1038/s41592-026-03236-1

Image Credits: AI Generated

DOI: 10.1038/s41592-026-03236-1

Keywords: spatial biology, VIMA, variational autoencoder, microniches, case-control analysis, rheumatoid arthritis, ulcerative colitis, dementia, MERFISH, CODEX, spatial transcriptomics, machine learning

The problem CellART attacks is fundamental to how modern spatial platforms work. Instruments such as 10x Genomics’ Xenium and Visium HD, Vizgen’s MERFISH and BGI’s Stereo-seq can now resolve gene expression at subcellular scale, but they pay a price for that resolution. Some capture only sparse transcript counts per spot, while others measure a limited panel of genes. In imaging-based assays, transcripts are detected as individual dots scattered across a tissue image, and deciding which dots belong to which cell is far from trivial. On array-based platforms like Visium HD, the data arrive in tiny bins that frequently straddle cell boundaries, so a single bin may contain RNA from two or three neighboring cells. Without a reliable way to draw cell boundaries and assign transcripts, the single-cell picture that biologists crave remains frustratingly blurred.

CellART, described in a Brief Communication published in Nature Computational Science on 5 October 2026, tackles both challenges at once. The framework performs cell segmentation and cell-type annotation simultaneously, fusing three complementary data streams: the staining images that accompany spatial assays, the spatial transcriptomics data themselves, and single-cell RNA sequencing references that catalog the expression signatures of known cell types. The key insight is that these modalities carry mutually reinforcing information. Histology images reveal where nuclei and membranes lie; transcript distributions show where gene products accumulate; and reference atlases constrain what combinations of cell types are biologically plausible in a given tissue. By integrating deep learning with probabilistic modeling, CellART lets each data source correct the weaknesses of the others.

Technically, the segmentation module builds on ideas from computer vision, drawing on architectures such as feature pyramid networks and residual networks that have proven their worth in object detection and image recognition. These networks learn to delineate cell boundaries from the staining images and the spatial patterns of transcript density. On top of that segmentation layer sits a probabilistic model in the spirit of deep generative approaches to single-cell transcriptomics, which assigns each segmented cell a cell-type label by comparing its expression profile with the scRNA-seq reference. This joint design means that segmentation errors and annotation errors can be caught and corrected in a single pass rather than compounding across separate pipeline stages, a common failure mode when researchers chain together independent tools.

The breadth of validation is one of the study’s most striking features. The team applied CellART to datasets spanning at least four distinct spatial platforms: Xenium, Visium HD, MERFISH and Stereo-seq, across tissues including human lung cancer, human breast cancer, human colorectal cancer, human skin, and mouse brain. On the Xenium 2.0 human lung dataset, where 10x Genomics provides a ground-truth segmentation derived from mixed-fluorescence imaging, CellART’s boundaries were quantified against established methods including Baysor, Cellpose and ProSeg using F1-scores and Jaccard indices. On Visium HD mouse brain data, where the recommended 16-micrometer bins do not align with cells in the histology image, CellART and Bin2Cell were compared by transcript coverage rate and by the correlation between cytoplasmic and nuclear expression profiles of individual cells, covering nearly sixty thousand cells in one benchmark alone.

Accuracy in cell typing proved equally important. In a human lung section profiled with both Xenium and post-Xenium Visium HD, CellART recovered consistent cell-type labels across the two platforms, suggesting that its annotations reflect genuine biology rather than platform-specific artifacts. On mouse brain data from all four platforms, the framework reconstructed the layered architecture of the cortex and reproduced the hippocampal organization documented in the Allen Mouse Brain Atlas, with the extracted cells correlating strongly with the single-cell reference. The team also measured computational efficiency by subsampling tissues of increasing size, and CellART remained tractable where some competing approaches struggled; one alternative method, TopACT, did not complete at full tissue size, a point the authors extrapolated rather than measured.

The biological payoff is vividly illustrated in cancer tissue. In the Xenium breast cancer dataset, CellART resolved ten cell types with high consistency between replicates, and its sharper boundaries prevented a classic pitfall: transcripts from neighboring cells bleeding into incorrectly drawn boundaries. Where a standard pipeline combining 10x segmentation with scANVI misassigned cells as tumor because ERBB2 transcripts from adjacent malignant cells drifted inside their borders, CellART produced crisper ERBB2 spatial patterns with less transcript diffusion. In breast cancer, where ERBB2 defines the HER2-positive subtype and guides therapy, that distinction between a tumor cell and its neighbor is not a cosmetic detail; it can change how a tumor’s molecular profile is read.

Perhaps the most compelling demonstration came from Visium HD colorectal cancer data, where CellART’s single-cell resolution outperformed spot-level deconvolution by RCTD, which classified the majority of more than 429,000 spots as doublets containing mixed cell populations. Working at true single-cell scale, CellART separated macrophage subpopulations that the coarser analysis blurred together. It distinguished SELENOP-positive macrophages scattered among tumor cells from SPP1-positive macrophages clustered at specific regions, one of which was recovered only by CellART. Defining tumor-associated macrophages as macrophages lying within fifty micrometers of a tumor cell, the team found these cells concentrating along tumor boundaries and forming a transcriptionally distinct cluster between macrophages and tumor cells, with upregulated immune-related and tumor-promoting genes, particularly MMP12.

That last finding hints at the kind of discovery CellART makes routine. Ligand-receptor analysis revealed strong interactions between tumor-associated macrophages and tumor cells, including the MMP12-PLAUR, SPP1-ITGAV and ITGB5, and APOE-SORL1 pairs, with their spatial expression co-localized at tumor margins. The urokinase plasminogen activator system, to which PLAUR belongs, has long been implicated in cancer metastasis and prognosis, and integrins are established drivers of invasion. Seeing these interaction axes light up precisely where macrophages and malignant cells touch provides a spatially grounded view of the tumor microenvironment that neither dissociated single-cell sequencing nor coarse spatial deconvolution can deliver. For immunotherapy research, where the dialogue between immune cells and tumor cells is everything, tools of this kind could reshape how tumor sections are interpreted.

Accessibility may determine whether CellART achieves the adoption its designers hope for. The software is written in Python, released under the permissive MIT licence on GitHub, and archived on Zenodo, with simulation and evaluation scripts, documentation and worked examples included. Its outputs are compatible with widely used community tools, including the SpatialData framework for spatial omics, easing integration into existing analysis workflows. All the spatial and single-cell reference datasets analyzed in the study are publicly available from 10x Genomics, Vizgen, BGI, the Allen Institute, the CZI CELLxGENE portal and the Allen Brain Atlas, and the processed inputs and simulated benchmark data are deposited for reuse. In a field where proprietary pipelines and incompatible file formats often fragment progress, a unified, open, platform-agnostic tool that turns raw high-resolution spatial data into trustworthy single-cell atlases could become as indispensable as the sequencing machines that generate the data in the first place.

Subject of Research: A unified computational framework for single-cell segmentation and annotation in high-resolution spatial transcriptomics

Article Title: CellART: a unified framework for extracting single-cell information from high-resolution spatial transcriptomics

Article References: CellART: a unified framework for extracting single-cell information from high-resolution spatial transcriptomics. (n.d.). https://doi.org/10.1038/s43588-026-01054-1

Image Credits: AI Generated

DOI: 10.1038/s43588-026-01054-1

Keywords: spatial transcriptomics, CellART, single-cell analysis, cell segmentation, deep learning, probabilistic modeling, tumor microenvironment, macrophages, Xenium, Visium HD, MERFISH, computational biology

The difficulty is geometric. Tissue slices are thin sections cut from three-dimensional structures, and the act of cutting, mounting, and preparing them introduces rotations, shifts, and stretches. One slice of a mouse brain may be tilted a few degrees relative to another; one may be slightly distorted so that its cortical layers sit at different distances from the edge. When researchers overlay two such maps, equivalent regions do not automatically coincide. Comparing gene activity in the wrong regions produces misleading conclusions, so meaningful comparison first requires alignment, a computational task known as registration, in which one digital map is shifted, rotated, and smoothly reshaped until its regions correspond to those of another.

Researchers at Kanazawa University in Japan and Sapienza University of Rome have now introduced a method designed to perform this registration directly on the raw spatial transcriptomics data, without sacrificing the cell-level resolution that makes the technology valuable. The approach, called Domain Elastic Transform, or DET, was developed by Osamu Hirose of Kanazawa University in collaboration with Emanuele Rodolà of Sapienza University of Rome and published online in IEEE Transactions on Pattern Analysis and Machine Intelligence on September 15, 2026. The work addresses a persistent trade-off in the field between preserving fine spatial detail and achieving robust alignment.

That trade-off arises because existing alignment strategies each make compromises. Image-based methods can incorporate gene-activity information, but they generally begin by converting the measurements into a regular grid of pixels, a step that blurs fine structures and reduces cell-level detail. Point-based methods, by contrast, keep the measurements as individual points and align them using their locations alone, but this can confuse regions whose shapes are similar while their gene-expression patterns differ. Other approaches can link cells or regions across samples without estimating how the tissue map should be smoothly reshaped, leaving the underlying geometric distortion unresolved. Each strategy uses only part of the information that spatial transcriptomics provides.

DET takes a different route by treating every measured cell or location as carrying two kinds of evidence at once: where it sits in the tissue and which genes are active there. The method represents each cell by its position and its measured gene-activity profile, then proceeds through an iterative process with two alternating steps. First, it estimates likely matches between cells in the two maps, drawing on both their spatial positions and the similarity of their gene-activity values. Second, it adjusts the positions of cells in one digital map, encouraging neighboring cells to move together so that the transformation remains smooth rather than chaotic. Repeating these steps gradually refines the alignment until it stabilizes.

Two design choices distinguish the method. DET modifies the positions of cells in the digital map but never alters the recorded gene-activity values themselves; the expression data serve only as clues about which regions should match. And the method requires no collection of pre-aligned training examples and no manually identified correspondences, making it training-free and unsupervised. In principle, this means it can be applied to any pair of tissue maps without the laborious preparation of reference alignments, a significant practical advantage in a field where verified ground truth is scarce.

The researchers evaluated DET against competing methods in 90 test cases using mouse brain maps generated with MERFISH, a technique that maps gene activity at single-cell resolution. To make the task deliberately difficult, they digitally rotated the maps by random angles spanning the full range from 0 to 360 degrees and shifted them by large distances. Among the methods tested, DET achieved the highest scores on three measures: how well the tissue maps overlapped after alignment, whether neighboring cells remained neighbors in the aligned result, and how closely gene-activity patterns agreed between corresponding regions. For the gene-activity comparison, the researchers combined expression information from nearby cells before comparing patterns, providing a robust assessment of functional correspondence.

The team then confronted a harder biological scenario: developing embryos, whose shapes change over time along with the types and proportions of cells they contain. Using maps from the MOSTA atlas, a collection of mouse development maps created with Stereo-seq, a technique that records gene activity across tissue, they aligned mouse-embryo maps from two developmental stages one day apart. Each map contained more than 100,000 measurement locations. DET first adjusted the overall position and orientation of the maps, then smoothly reshaped one to refine the alignment. Because no verified reference existed showing exactly which locations should match across these stages, the experiment served primarily to demonstrate that the method could be applied to maps of this size and complexity, a scale at which many alignment approaches struggle.

Scalability was tested further by artificially expanding MERFISH-based data to as many as one million points per tissue slice. Peak memory use averaged less than one gigabyte in every setting tested. The method achieves this efficiency through a landmark strategy: it estimates the alignment using a selected subset of points, called landmarks, and extends the estimated movement to the full map. The number of landmarks, which the researchers call the landmark budget, controls how much computation the reshaping step requires. With the number of landmarks held fixed, peak memory use grew approximately in proportion to the total number of points, while computation time depended mainly on the number of landmarks used in the main alignment step rather than on the full dataset size. The researchers note that exact runtimes depend on the computer and algorithm settings, but the results support applying DET to very large tissue maps.

The implications extend beyond tissue slices. Many scientific datasets combine locations with measurements, and similar alignment challenges arise with data attached to three-dimensional surfaces or anatomical structures. The researchers position DET as a complement to existing alignment methods rather than a universal replacement, designed for tasks that require both fine spatial detail and a smooth transformation of the map rather than only a list of matching locations. The broader aim, as Hirose describes it, is to use both kinds of information that such datasets carry, where something is and what was measured there, when bringing different datasets into alignment, rather than treating shape and measurements separately. For spatial transcriptomics, that could mean more reliable comparisons of corresponding regions across samples, even when their shapes and gene-activity patterns differ, strengthening studies of how organs develop and how disease reorganizes the body’s molecular geography. The research was supported in part by the Japan Science and Technology Agency FOREST Program under grant JPMJFR242V.

Subject of Research: A computational registration method for aligning spatial transcriptomics maps of gene activity across tissue samples

Article Title: Making maps of gene activity easier to compare

Article References: Making maps of gene activity easier to compare. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: spatial transcriptomics, image registration, Domain Elastic Transform, gene expression, MERFISH, Stereo-seq, mouse brain, mouse embryo, MOSTA atlas, computational biology, Kanazawa University, IEEE TPAMI

The technical centerpiece of the study is a method the team calls STICR, a lentiviral barcoding system that stamps progenitor cells with unique heritable DNA sequences. When a barcoded progenitor divides, all of its descendants carry the same genetic tag, allowing researchers to reconstruct family trees long after development has moved on. By injecting the virus at a series of embryonic and postnatal timepoints—from embryonic day 11 through postnatal day 0—and collecting tissue at postnatal day 4, the team captured clones born during distinct developmental windows. Each labeled cell was then profiled with single-cell RNA sequencing, and its barcode was read out alongside its transcriptome, linking identity to ancestry in a single measurement.

Knowing what a cell is, however, is only half the story; knowing where it sits is equally important. To anchor their lineage data in anatomy, the researchers used MERFISH, a spatial transcriptomics method that maps thousands of genes across intact coronal sections of the forebrain. By registering the cellular profiles recovered from dissociated tissue onto these spatial maps, the team could ask not only which cell types share a common ancestor, but also how far those relatives traveled from one another. The result is a spatiotemporal clonal atlas: a record of lineage, identity and location woven together across the entire newborn forebrain rather than a single privileged region such as the cortex, which has dominated most prior lineage studies.

The scale of the dataset allowed the authors to run careful computational checks on their inferences. Because viral labeling captures only a fraction of cells, the team built simulation frameworks using ground-truth clones to model how partial sampling affects the recovery of lineage structure. These simulations, described in the paper’s extended data, quantified how many multicellular clones are needed to reliably detect the constituent cell types of a lineage and how often apparent fate biases could arise by chance. Such validation matters, because the study’s headline conclusions rest on statistical patterns of clone sharing between cell types rather than on direct observation of dividing cells.

Among the most consequential findings is a new view of the striatum, the brain’s major input hub for movement and reward. The atlas reveals previously unrecognized differences in how ventral and dorsal medium spiny neurons—the principal projection neurons of the striatum—are generated. It also uncovers lineage relationships that had gone unexplored: late-born medium spiny neurons share clonal ancestry with striatal astrocytes and with distinct subpopulations of olfactory bulb neurons. This connects the striatum’s developmental program to the continuous postnatal production of neurons that migrate into the olfactory bulb, a hallmark of the subventricular zone that has long fascinated neuroscientists studying adult neurogenesis.

The broader principle that emerges from the atlas is a tale of two lineages. Glutamatergic neurons—the forebrain’s excitatory workhorses—and astrocytes, the star-shaped glial cells that support synapses, are clonally related to one another and remain within their region of origin. A clone born in the cortex stays in the cortex; one born in the hippocampus stays in the hippocampus. GABAergic neurons, the inhibitory counterpart, follow a different rule. They are clonally linked to oligodendrocyte precursor cells, the myelin-forming lineage, and they disperse extensively, crossing regional and structural boundaries. The findings echo and extend earlier work showing that clonally related interneurons spread widely across the forebrain, but the new atlas places that dispersal within a whole-brain framework and ties it to glial fate.

That dispersal carries a provocative clinical hint. The researchers found that clones of GABAergic neurons with broad regional dispersion are enriched for genes implicated in neurodevelopmental disorders, including genes associated with autism. In other words, the cell lineages most likely to spread across the brain are also the ones whose genetic programs intersect most heavily with disease risk. This observation suggests that disruptions to widely dispersing inhibitory lineages could have consequences distributed across multiple brain regions, offering a potential developmental framework for understanding why genetic risk factors for neurodevelopmental conditions produce such varied and widespread effects.

The temporal dimension of the atlas adds further texture. By comparing clones labeled at different embryonic and postnatal stages, the team traced how lineage composition shifts over time: early-labeled clones are dominated by deep-layer cortical neurons, while later windows capture upper-layer neurons, glia and the postnatally generated populations destined for the olfactory bulb. In the olfactory bulb itself, the data distinguish embryonically or locally generated GABAergic neurons from postnatally born neurons still migrating through the rostral migratory stream at the time of collection. Cajal–Retzius cells, the pioneer neurons that guide cortical layering, also appear in the atlas with their own regionally specific lineage relationships to astrocytes and ependymal cells.

Methodologically, the study demonstrates how viral barcoding, single-cell transcriptomics and spatial transcriptomics can be fused into a single coherent picture of organ-scale development. The authors have made their resources broadly available: the single-cell and barcode datasets are deposited in the Gene Expression Omnibus under accession GSE341480, an interactive browser of the single-cell and MERFISH data is hosted through the UCSC cell browser, the image dataset is available through the Brain Image Library, and the custom analysis code is published on GitHub. For a field increasingly reliant on large-scale atlases, this openness means other laboratories can immediately interrogate the lineage relationships, test alternative hypotheses and integrate the data with their own developmental and disease-focused studies.

The implications reach well beyond the mouse. Human studies have suggested that individual cortical progenitors can produce both excitatory and inhibitory neurons, and the new mouse atlas provides a mechanistic reference against which human lineage data can be interpreted. By establishing that lineage fate—whether a progenitor’s descendants become neurons, astrocytes or oligodendrocyte precursor cells—is governed by regionally specific rules that differ fundamentally between excitatory and inhibitory lineages, the work offers a foundational framework for dissecting how the mammalian forebrain is assembled, and how that assembly goes awry in disease. As the authors note, the atlas is intended as a reference point for the field: a first whole-forebrain census of who is related to whom, where they ended up, and when they were born.

Subject of Research: Clonal lineage relationships and cell fate specification during mouse forebrain development

Article Title: Spatiotemporal clonal architecture of the newborn mouse forebrain

Article References: Yuan, G., Kunst, M., Steyert, M. R., Keefe, M. G., Mathieu, R., van Velthoven, C. T. J., McMillen, D., Waters, J., Fukushima, Y., Kazerounian, A., Alvarez-Buylla, A., Zeng, H., & Nowakowski, T. J. (2026). Spatiotemporal clonal architecture of the newborn mouse forebrain. Nature. https://doi.org/10.1038/s41586-026-11064-8

Image Credits: AI Generated

DOI: 10.1038/s41586-026-11064-8

Keywords: mouse forebrain, clonal lineage, lentiviral barcoding, spatial transcriptomics, MERFISH, single-cell RNA sequencing, GABAergic neurons, glutamatergic neurons, astrocytes, oligodendrocyte precursor cells, neurodevelopmental disorders, striatum