Review

visium spatial transcriptome data (10X Genomics)

About

News

Press Release

Team

Advisors

Partners

Contact

Bioz Stars

Bioz vStars

Buy from Supplier

Structured Review

10X Genomics visium spatial transcriptome data

Schema representation of STaCker. The workflow ( a ) takes as inputs the tissue images from a pair of reference and moving spatial <t>transcriptome</t> slices, combined with the contour maps generated upon the gene expression profiles from the respective slices. The resulting composite images are subsequently aligned through a deep neural network-based registration module. The registration module outputs the inferred deformation field to align the spatial coordinates of the spots/cells in the moving slice. The architecture of the registration module ( b ) takes a four-level contracting path and a four-level expanding path with skip connections at all levels. The final layer of the decoder is further convoluted to generate the spatial velocity field followed by a vector integration to output the deformation field for the alignment. Synthetic images with segmentation label maps (Methods) are used to train the module. The moved label map, after applying the deformation field to the moving label map, is compared to the reference label map. The difference constitutes the key component in the loss function.

Visium Spatial Transcriptome Data, supplied by 10X Genomics, used in various techniques. Bioz Stars score: 90/100, based on 1 PubMed citations. ZERO BIAS - scores, article reviews, protocol conditions and more
https://www.bioz.com/product/spatial+transcriptomics+visium+data/pmc12103625-540-1-19?v=10X+Genomics
Average 90 stars, based on 1 article reviews

visium spatial transcriptome data - by Bioz Stars, 2026-07

90/100 stars

Images

1) Product Images from "Image guided construction of a common coordinate framework for spatial transcriptome data"

Article Title: Image guided construction of a common coordinate framework for spatial transcriptome data

Journal: Scientific Reports

doi: 10.1038/s41598-025-01862-x

Schema representation of STaCker. The workflow ( a ) takes as inputs the tissue images from a pair of reference and moving spatial transcriptome slices, combined with the contour maps generated upon the gene expression profiles from the respective slices. The resulting composite images are subsequently aligned through a deep neural network-based registration module. The registration module outputs the inferred deformation field to align the spatial coordinates of the spots/cells in the moving slice. The architecture of the registration module ( b ) takes a four-level contracting path and a four-level expanding path with skip connections at all levels. The final layer of the decoder is further convoluted to generate the spatial velocity field followed by a vector integration to output the deformation field for the alignment. Synthetic images with segmentation label maps (Methods) are used to train the module. The moved label map, after applying the deformation field to the moving label map, is compared to the reference label map. The difference constitutes the key component in the loss function.

Figure Legend Snippet: Schema representation of STaCker. The workflow ( a ) takes as inputs the tissue images from a pair of reference and moving spatial transcriptome slices, combined with the contour maps generated upon the gene expression profiles from the respective slices. The resulting composite images are subsequently aligned through a deep neural network-based registration module. The registration module outputs the inferred deformation field to align the spatial coordinates of the spots/cells in the moving slice. The architecture of the registration module ( b ) takes a four-level contracting path and a four-level expanding path with skip connections at all levels. The final layer of the decoder is further convoluted to generate the spatial velocity field followed by a vector integration to output the deformation field for the alignment. Synthetic images with segmentation label maps (Methods) are used to train the module. The moved label map, after applying the deformation field to the moving label map, is compared to the reference label map. The difference constitutes the key component in the loss function.

Techniques Used: Generated, Gene Expression, Plasmid Preparation

Evaluation of STaCker in aligning digitally warped spatial transcriptome slices of mouse brain. ( a ) The reference is a mouse sagittal posterior brain slice profiled by 10 × Genomics Visium platform. It was digitally warped using Simplex noises to a low (noise amplitude = 5, NCC of the deformed image = 0.61), medium (noise amplitude = 10, NCC of the deformed image = 0.57), or high (noise amplitude = 15, NCC of the deformed image = 0.54 level to generate a series of moving slices (noise frequency remains 1 for all warping). ( b ) The discordance between the spatial coordinates of the spots in each moving slice and those in the reference slice is quantified by the MSE (Methods) shown in the bar plots. STaCker, together with previously published methods STUtility, PASTE, and GPSA, was applied to align each of the moving spatial transcriptome slices to the reference. The spots’ coordinates before (blue crosses) or after the alignment (red crosses) are displayed together with the reference spot coordinates (gray dots) to aid the visual comparison. The post-alignment MSEs from each method are illustrated in the bar plots. Value from STaCker is the mean over five runs, shown together with the standard errors as error bars. STaCker’s MSE is significantly lower than that of all other programs (one sample t-test p -values < = 1e-3).

Figure Legend Snippet: Evaluation of STaCker in aligning digitally warped spatial transcriptome slices of mouse brain. ( a ) The reference is a mouse sagittal posterior brain slice profiled by 10 × Genomics Visium platform. It was digitally warped using Simplex noises to a low (noise amplitude = 5, NCC of the deformed image = 0.61), medium (noise amplitude = 10, NCC of the deformed image = 0.57), or high (noise amplitude = 15, NCC of the deformed image = 0.54 level to generate a series of moving slices (noise frequency remains 1 for all warping). ( b ) The discordance between the spatial coordinates of the spots in each moving slice and those in the reference slice is quantified by the MSE (Methods) shown in the bar plots. STaCker, together with previously published methods STUtility, PASTE, and GPSA, was applied to align each of the moving spatial transcriptome slices to the reference. The spots’ coordinates before (blue crosses) or after the alignment (red crosses) are displayed together with the reference spot coordinates (gray dots) to aid the visual comparison. The post-alignment MSEs from each method are illustrated in the bar plots. Value from STaCker is the mean over five runs, shown together with the standard errors as error bars. STaCker’s MSE is significantly lower than that of all other programs (one sample t-test p -values < = 1e-3).

Techniques Used: Slice Preparation, Comparison

Performance of STaCker in the de novo alignment of spatial transcriptome slices. The top row displays four moving spatial transcriptome slices that were independently warped from a reference slice taken from the mouse posterior brain used in Fig. , with the spot coordinates shown as crosses over the tissue images (slice 1: red, slice 2: green, slice 3: blue, slice 4: orange). The warping was conducted using random-seeded Simplex noises with an amplitude of 15 and a frequency of 1. The mean pairwise NCCs among the tissue images of the moving slices is 0.198. The average pairwise MSE among the spot coordinates in the moving slices is 0.10. The bottom row illustrates the spot coordinates from four slices before the alignment (“Unaligned coordinates”) and after the alignment by STaCker, STUtility, PASTE, GPSA, respectively, using the same colors and cross symbols as shown in the top row. The post-alignment average MSE over all six pairs of slices is 0.043, 0.119, 0.098, 0.601 for STaCker, STUtility, PASTE, and GPSA, respectively.

Figure Legend Snippet: Performance of STaCker in the de novo alignment of spatial transcriptome slices. The top row displays four moving spatial transcriptome slices that were independently warped from a reference slice taken from the mouse posterior brain used in Fig. , with the spot coordinates shown as crosses over the tissue images (slice 1: red, slice 2: green, slice 3: blue, slice 4: orange). The warping was conducted using random-seeded Simplex noises with an amplitude of 15 and a frequency of 1. The mean pairwise NCCs among the tissue images of the moving slices is 0.198. The average pairwise MSE among the spot coordinates in the moving slices is 0.10. The bottom row illustrates the spot coordinates from four slices before the alignment (“Unaligned coordinates”) and after the alignment by STaCker, STUtility, PASTE, GPSA, respectively, using the same colors and cross symbols as shown in the top row. The post-alignment average MSE over all six pairs of slices is 0.043, 0.119, 0.098, 0.601 for STaCker, STUtility, PASTE, and GPSA, respectively.

Techniques Used:

Coordinate consolidation of real spatial transcriptome slices from a human dorsolateral prefrontal cortex (DLPFC). ( a ): Four serial dissections of the dorsolateral prefrontal cortex. ( b ): Superimposed spatial coordinates of the QC-validated tissue spots from four DLPFC slices before and after alignment by different methods, color-coded by their tissue domain annotations in the original publication. ( c - d ): Quantitative evaluation of tissue domain consistency across the four slices before and after alignment, using Spatial Coherence Score ( c ) and Mean Pairwise Adjusted Rand Index ( d ). STUtility does not offer de novo alignments so their values are averaged over four alignments, each using a different slice as the fixed template, with error bars marking the standard errors. ( e ): Spatial patterns of representative genes before and after alignment by STaCker and other programs. The displayed expression values are the natural logarithm transformation of the normalized UMI count (10 4 total UMI counts per spot). ( f ): Comparison of Moran’s I spatial autocorrelations of representative genes before and after alignment by various programs.

Figure Legend Snippet: Coordinate consolidation of real spatial transcriptome slices from a human dorsolateral prefrontal cortex (DLPFC). ( a ): Four serial dissections of the dorsolateral prefrontal cortex. ( b ): Superimposed spatial coordinates of the QC-validated tissue spots from four DLPFC slices before and after alignment by different methods, color-coded by their tissue domain annotations in the original publication. ( c - d ): Quantitative evaluation of tissue domain consistency across the four slices before and after alignment, using Spatial Coherence Score ( c ) and Mean Pairwise Adjusted Rand Index ( d ). STUtility does not offer de novo alignments so their values are averaged over four alignments, each using a different slice as the fixed template, with error bars marking the standard errors. ( e ): Spatial patterns of representative genes before and after alignment by STaCker and other programs. The displayed expression values are the natural logarithm transformation of the normalized UMI count (10 4 total UMI counts per spot). ( f ): Comparison of Moran’s I spatial autocorrelations of representative genes before and after alignment by various programs.

Techniques Used: Expressing, Transformation Assay, Comparison

Coordinate consolidation of real spatial transcriptome slices from independent replicates of mouse olfactory bulbs. ( a ): Four biological replicates of dissected mouse olfactory bulbs. ( b ): Superimposed spatial coordinates of the QC-validated tissue spots from four slices before and after alignment by different methods, color-coded by the tissue domain annotations derived upon the transcriptome of the spots. ( c - d ): Quantitative evaluation of tissue domain consistency across the four slices before and after alignment, using Spatial Coherence Score ( c ) and Mean Pairwise Adjusted Rand Index ( d ). STUtility does not offer de novo alignments so their values are averaged over four alignments, each using a different slice as the fixed template. The standard errors are shown as error bars. ( e ): Spatial patterns of representative genes before and after alignment by STaCker and the other programs. The displayed expression is after the natural logarithm transformation of the normalized UMI count (10 4 total UMI counts per spot). ( f ): Comparison of Moran’s I spatial autocorrelations of representative genes before and after alignment by various programs.

Figure Legend Snippet: Coordinate consolidation of real spatial transcriptome slices from independent replicates of mouse olfactory bulbs. ( a ): Four biological replicates of dissected mouse olfactory bulbs. ( b ): Superimposed spatial coordinates of the QC-validated tissue spots from four slices before and after alignment by different methods, color-coded by the tissue domain annotations derived upon the transcriptome of the spots. ( c - d ): Quantitative evaluation of tissue domain consistency across the four slices before and after alignment, using Spatial Coherence Score ( c ) and Mean Pairwise Adjusted Rand Index ( d ). STUtility does not offer de novo alignments so their values are averaged over four alignments, each using a different slice as the fixed template. The standard errors are shown as error bars. ( e ): Spatial patterns of representative genes before and after alignment by STaCker and the other programs. The displayed expression is after the natural logarithm transformation of the normalized UMI count (10 4 total UMI counts per spot). ( f ): Comparison of Moran’s I spatial autocorrelations of representative genes before and after alignment by various programs.

Techniques Used: Derivative Assay, Expressing, Transformation Assay, Comparison

Alignment of ISH-based spatial transcriptome slices. ( a ) MERFISH slices from three mouse brain samples, illustrated using the DAPI staining images. ( b ) Superimposed spatial coordinates of cells in the three slices before and after alignment by different methods. Cells are color-coded by their niches, defined based on the gene profiling of the cells (see Methods). ( c ) Quantitative evaluation of tissue domain consistency across the three slices before and after alignment, using Spatial Coherence Score and Mean Pairwise Adjusted Rand Index. STUtility and STalign do not offer de novo alignments so their values are averaged over three alignments each with a different slice as the fixed template, with standard errors shown as error bars. ( d ) Spatial patterns of four representative genes before and after alignment by STaCker, STUtility and STalign. The displayed expression is after the natural logarithm transformation of the normalized UMI count (10 4 total UMI counts per cell). ( e ) Comparison of Moran’s I autocorrelation (left panel) and Gene Coherence Score (right panel) of the four representative genes from the slices aligned by STaCker, STUtility or STalign. Values from STUtility and STalign are the average over three alignments each using a different slice as the fixed template, with standard errors shown as error bars. ( f ) Boxplots of the Moran’s I autocorrelation (left panel) and Gene Coherence Score (right panel) of all non-randomly distributed genes (Moran’s I p -value < = 0.01) over the slices aligned by STaCker, STUtility or STalign. The top and bottom edges of the box represent the 3rd and 1st quantiles, with the horizontal line inside denoting the median. The ends of the whisker mark the 1.5 times interquartile range, calculated as the difference between the 3rd and 1st quartiles, from the box edges. Data points beyond the whisker range are represented as dots. STalign was executed using the same parameters applied to the same MERFISH dataset in the original publication. ( g ) Comparison of the spatial coherence in gene expressions after the alignment by STaCker and STalign. Genes that show significantly higher Moran’s I correlation (left panel) or Gene Coherence Score (GCS, right panel) after alignment with STaCker are marked with orange dots, while genes with significantly elevated values for these metrics following STalign alignment are indicated by blue dots. The dashed line represents the significance cutoff (0.05) for the two-sided Student’s t-test p -value.

Figure Legend Snippet: Alignment of ISH-based spatial transcriptome slices. ( a ) MERFISH slices from three mouse brain samples, illustrated using the DAPI staining images. ( b ) Superimposed spatial coordinates of cells in the three slices before and after alignment by different methods. Cells are color-coded by their niches, defined based on the gene profiling of the cells (see Methods). ( c ) Quantitative evaluation of tissue domain consistency across the three slices before and after alignment, using Spatial Coherence Score and Mean Pairwise Adjusted Rand Index. STUtility and STalign do not offer de novo alignments so their values are averaged over three alignments each with a different slice as the fixed template, with standard errors shown as error bars. ( d ) Spatial patterns of four representative genes before and after alignment by STaCker, STUtility and STalign. The displayed expression is after the natural logarithm transformation of the normalized UMI count (10 4 total UMI counts per cell). ( e ) Comparison of Moran’s I autocorrelation (left panel) and Gene Coherence Score (right panel) of the four representative genes from the slices aligned by STaCker, STUtility or STalign. Values from STUtility and STalign are the average over three alignments each using a different slice as the fixed template, with standard errors shown as error bars. ( f ) Boxplots of the Moran’s I autocorrelation (left panel) and Gene Coherence Score (right panel) of all non-randomly distributed genes (Moran’s I p -value < = 0.01) over the slices aligned by STaCker, STUtility or STalign. The top and bottom edges of the box represent the 3rd and 1st quantiles, with the horizontal line inside denoting the median. The ends of the whisker mark the 1.5 times interquartile range, calculated as the difference between the 3rd and 1st quartiles, from the box edges. Data points beyond the whisker range are represented as dots. STalign was executed using the same parameters applied to the same MERFISH dataset in the original publication. ( g ) Comparison of the spatial coherence in gene expressions after the alignment by STaCker and STalign. Genes that show significantly higher Moran’s I correlation (left panel) or Gene Coherence Score (GCS, right panel) after alignment with STaCker are marked with orange dots, while genes with significantly elevated values for these metrics following STalign alignment are indicated by blue dots. The dashed line represents the significance cutoff (0.05) for the two-sided Student’s t-test p -value.

Techniques Used: Staining, Expressing, Transformation Assay, Comparison, Whisker Assay

Alignment across different spatial transcriptome platforms. ( a ) Two mouse brain hemispheres profiled using 10 × Genomics Visium and Xenium, shown as the acquired H&E and DAPI images, respectively. ( b ) Superimposed spatial coordinates of spots (red circles, Visium slice) or cells (blue crosses, Xenium slice) before and after alignment by different methods. ( c ) Spatial patterns of the representative genes before and after alignment by STaCker, STUtility or STalign. The positions of the spots in the Visium slice (red) and the 55-micron × 55-micron pseudo-spots in the Xenium slice (blue) are displayed. At each spot or pseudo-spot, the expression of a gene is divided by the maximum expression of that gene on the slice, converting it to a value within 0 and 1. The scaled gene expressions are comparable across platforms and thus used for visualization and quantitative evaluation. ( d ) Moran’s I autocorrelation (upper panel) and the Gene Coherence Score (lower panel) of the representative genes from the slices aligned by STaCker, STUtility or STalign. Values for STUtility and STalign that do not offer de novo alignment are the average over alignments each using one of the slices as the reference, with the standard errors shown as error bars. ( e ) Boxplots of the Moran’s I autocorrelation score (upper panel) and the Gene Coherence Score (lower panel) of all non-randomly distributed genes (Moran’s I p -value < = 0.01) over the slices aligned by STaCker, STUtility or STalign. For both metrics, the mean of the distribution in STaCker is significantly higher than that in STalign (two-sided student t-test p -value < = 6e-5) and in STUtility (two-sided student t-test p -value < = 2e-3). In all boxplots, the top and bottom edges of the box represent the 3rd and 1st quantiles with the horizontal line inside to denote the median. The whiskers extend to 1.5 times the interquartile range (IQR), which is the difference between the 3rd and 1st quantiles, from the box edges. Data points outside the whisker range are displayed as dots. ( f ) Comparison of spatial coherence in gene expressions following alignment using STaCker and STalign. Genes exhibiting significantly higher Moran’s I correlation (left panel) or Gene Coherence Score (right panel) after alignment with STaCker are depicted with orange dots. Genes with significantly increased values for the two metrics after alignment with STalign are shown with blue dots. The dashed line indicates the significance cutoff (0.05) for the two-sided Student’s t-test p -value.

Figure Legend Snippet: Alignment across different spatial transcriptome platforms. ( a ) Two mouse brain hemispheres profiled using 10 × Genomics Visium and Xenium, shown as the acquired H&E and DAPI images, respectively. ( b ) Superimposed spatial coordinates of spots (red circles, Visium slice) or cells (blue crosses, Xenium slice) before and after alignment by different methods. ( c ) Spatial patterns of the representative genes before and after alignment by STaCker, STUtility or STalign. The positions of the spots in the Visium slice (red) and the 55-micron × 55-micron pseudo-spots in the Xenium slice (blue) are displayed. At each spot or pseudo-spot, the expression of a gene is divided by the maximum expression of that gene on the slice, converting it to a value within 0 and 1. The scaled gene expressions are comparable across platforms and thus used for visualization and quantitative evaluation. ( d ) Moran’s I autocorrelation (upper panel) and the Gene Coherence Score (lower panel) of the representative genes from the slices aligned by STaCker, STUtility or STalign. Values for STUtility and STalign that do not offer de novo alignment are the average over alignments each using one of the slices as the reference, with the standard errors shown as error bars. ( e ) Boxplots of the Moran’s I autocorrelation score (upper panel) and the Gene Coherence Score (lower panel) of all non-randomly distributed genes (Moran’s I p -value < = 0.01) over the slices aligned by STaCker, STUtility or STalign. For both metrics, the mean of the distribution in STaCker is significantly higher than that in STalign (two-sided student t-test p -value < = 6e-5) and in STUtility (two-sided student t-test p -value < = 2e-3). In all boxplots, the top and bottom edges of the box represent the 3rd and 1st quantiles with the horizontal line inside to denote the median. The whiskers extend to 1.5 times the interquartile range (IQR), which is the difference between the 3rd and 1st quantiles, from the box edges. Data points outside the whisker range are displayed as dots. ( f ) Comparison of spatial coherence in gene expressions following alignment using STaCker and STalign. Genes exhibiting significantly higher Moran’s I correlation (left panel) or Gene Coherence Score (right panel) after alignment with STaCker are depicted with orange dots. Genes with significantly increased values for the two metrics after alignment with STalign are shown with blue dots. The dashed line indicates the significance cutoff (0.05) for the two-sided Student’s t-test p -value.

Techniques Used: Expressing, Whisker Assay, Comparison

Image Search Results

Comparative spatial multi-omics analysis of acute myeloid leukemia patients’ bone marrow and extramedullary tissues (A) Schematic representation of the study workflow. Paired bone marrow (BM) samples (BM1 and BM2) and extramedullary (EM) samples (EM1, from skin; and EM2, from lymph node) from 2 newly diagnosed patients with acute myeloid leukemia (AML) (PT1 and PT2) were fixed in formalin and embedded in paraffin (FFPE) and then sectioned for use in Visium assays (v1 and v2), and Opal multiplex fluorescent immunohistochemistry (mfIHC). The Visium spatial transcriptomics (ST) results were validated using GeoMx digital spatial profiling (DSP) with tissue microarrays (TMAs) of samples from 3 newly diagnosed patients with AML (PT3, PT4, and PT5). An additional 4 AML bone marrow samples that performed the Visium gene and protein expression assay are used as a validational cohort (PT6, PT7, PT8, and PT9). Image created with BioRender ( https://biorender.com ). (B) Uniform manifold approximation and projection (UMAP) plot showing our reference map consisted of 79,029 cells collected from 9 healthy BM donors and 7 patients with AML with diploid cytogenetics to match the patient cytogenetic profiles, and included both newly generated scRNA data and previous works. This map consisted of 21 cell types, including T cells (CD4 + and CD8 + naive, effector, and memory T cells, T regulatory [Treg] cells, and unconventional T cells), other immune cells (Natural killer [NK] cells, B cells and plasma cells), hematopoietic progenitors (Hematopoietic stem cells [HSCs], common lymphoid progenitors [CLPs], granulocyte-monocyte progenitors [GMPs]), myeloid cells (megakaryocytes/platelets, monocytes, early and late erythroid cells, conventional and plasmacytoid dendritic cells) and leukemic (AML) cell populations. (C) Immunohistochemical staining of CD11c, MPO, and CD3e on BM1 sections that were used for histopathological annotation. The scale bar for the main tissue panels represents 1 mm. The scale bar for the zoomed-in panels, corresponding to the boxed regions, represents 100 μm. (D) Unsupervised clustering and pathology annotation for the projected spatial map of BM1, revealing 3 distinct regions with an adjusted rand index (ARI) of 0.46. (E) Spatial deconvolution of BM1 tissue, showing erythroid and AML cell populations, with CD11c immunohistochemistry (IHC) overlaid on an image of hematoxylin and eosin (H&E)-staining. The dotted red lines represent regions enriched for the erythroid cell population; dotted black lines, regions enriched for the AML cell population; and solid lines, regions that overlapped with other tissue sections. (F) Heatmap of Z score normalized canonical markers in pathology annotations, with matching unsupervised cluster distributions represented as a pie chart. HBB, HBD, HBA2, GATA1/2 are erythroid genes and S100A12, FCGR3A, CD14, MS4A7, and , CD33 are monocyte/leukemic genes. (G) Representative overlay of Visium H&E staining with Opal mfIHC and the generated spot-level data for CD33, CD71, CXCL12, CXCR4, CD68, and IL-6. Boxes illustrate magnified regions showing concordance between transcript-level (Visium) and protein-level (Opal) signals at the spot level. (H) Phenotype staining on near-adjacent tissue sections for markers of leukemic (CD33), monocytic (CD68), and erythroid (CD71) populations. DAPI was used as a nuclear counterstain. The spatial distribution of these markers corroborates ST-based spot deconvolution. Scale bars: 1 mm (whole-slide panels) and 100 μm (selected region panels). (I) Box and spatial plots of mfIHC staining intensities for phenotypic markers across ST-defined clusters in BM1, highlighting the enrichment of leukemic and monocytic populations in cluster 3 and that of erythroid populations in cluster 2 at BM1. Scale bars: 1 mm (whole-slide panels) and 100 μm (selected region panels). ns, not significant. ∗∗∗∗ p < 0.0001, Wilcoxon rank-sum test.

Journal: iScience

Article Title: Integrative spatial multi-omics reveal niche-specific inflammatory signaling and differentiation hierarchies in AML

doi: 10.1016/j.isci.2025.114289

Figure Lengend Snippet: Comparative spatial multi-omics analysis of acute myeloid leukemia patients’ bone marrow and extramedullary tissues (A) Schematic representation of the study workflow. Paired bone marrow (BM) samples (BM1 and BM2) and extramedullary (EM) samples (EM1, from skin; and EM2, from lymph node) from 2 newly diagnosed patients with acute myeloid leukemia (AML) (PT1 and PT2) were fixed in formalin and embedded in paraffin (FFPE) and then sectioned for use in Visium assays (v1 and v2), and Opal multiplex fluorescent immunohistochemistry (mfIHC). The Visium spatial transcriptomics (ST) results were validated using GeoMx digital spatial profiling (DSP) with tissue microarrays (TMAs) of samples from 3 newly diagnosed patients with AML (PT3, PT4, and PT5). An additional 4 AML bone marrow samples that performed the Visium gene and protein expression assay are used as a validational cohort (PT6, PT7, PT8, and PT9). Image created with BioRender ( https://biorender.com ). (B) Uniform manifold approximation and projection (UMAP) plot showing our reference map consisted of 79,029 cells collected from 9 healthy BM donors and 7 patients with AML with diploid cytogenetics to match the patient cytogenetic profiles, and included both newly generated scRNA data and previous works. This map consisted of 21 cell types, including T cells (CD4 + and CD8 + naive, effector, and memory T cells, T regulatory [Treg] cells, and unconventional T cells), other immune cells (Natural killer [NK] cells, B cells and plasma cells), hematopoietic progenitors (Hematopoietic stem cells [HSCs], common lymphoid progenitors [CLPs], granulocyte-monocyte progenitors [GMPs]), myeloid cells (megakaryocytes/platelets, monocytes, early and late erythroid cells, conventional and plasmacytoid dendritic cells) and leukemic (AML) cell populations. (C) Immunohistochemical staining of CD11c, MPO, and CD3e on BM1 sections that were used for histopathological annotation. The scale bar for the main tissue panels represents 1 mm. The scale bar for the zoomed-in panels, corresponding to the boxed regions, represents 100 μm. (D) Unsupervised clustering and pathology annotation for the projected spatial map of BM1, revealing 3 distinct regions with an adjusted rand index (ARI) of 0.46. (E) Spatial deconvolution of BM1 tissue, showing erythroid and AML cell populations, with CD11c immunohistochemistry (IHC) overlaid on an image of hematoxylin and eosin (H&E)-staining. The dotted red lines represent regions enriched for the erythroid cell population; dotted black lines, regions enriched for the AML cell population; and solid lines, regions that overlapped with other tissue sections. (F) Heatmap of Z score normalized canonical markers in pathology annotations, with matching unsupervised cluster distributions represented as a pie chart. HBB, HBD, HBA2, GATA1/2 are erythroid genes and S100A12, FCGR3A, CD14, MS4A7, and , CD33 are monocyte/leukemic genes. (G) Representative overlay of Visium H&E staining with Opal mfIHC and the generated spot-level data for CD33, CD71, CXCL12, CXCR4, CD68, and IL-6. Boxes illustrate magnified regions showing concordance between transcript-level (Visium) and protein-level (Opal) signals at the spot level. (H) Phenotype staining on near-adjacent tissue sections for markers of leukemic (CD33), monocytic (CD68), and erythroid (CD71) populations. DAPI was used as a nuclear counterstain. The spatial distribution of these markers corroborates ST-based spot deconvolution. Scale bars: 1 mm (whole-slide panels) and 100 μm (selected region panels). (I) Box and spatial plots of mfIHC staining intensities for phenotypic markers across ST-defined clusters in BM1, highlighting the enrichment of leukemic and monocytic populations in cluster 3 and that of erythroid populations in cluster 2 at BM1. Scale bars: 1 mm (whole-slide panels) and 100 μm (selected region panels). ns, not significant. ∗∗∗∗ p < 0.0001, Wilcoxon rank-sum test.

Article Snippet: Visium Spatial Transcriptomics Data , This paper , GEO: GSE279576.

Techniques: Biomarker Discovery, Multiplex Assay, Immunohistochemistry, Expressing, Generated, Clinical Proteomics, Immunohistochemical staining, Staining

Spatial multi-omics profiling identifies leukemic infiltration and tissue composition in extramedullary acute myeloid leukemia samples (A) Unsupervised clustering of the extramedullary sample EM1 into 3 spatial clusters (left) compared against the pathology-based annotation (right; indicating a composition of leukemia, dermis, epidermis, and gland). The adjusted rand index (ARI; 0.51) reflects moderate agreement between the clusters and pathology annotations. (B) Spatial deconvolution scores obtained using the SpaCET algorithm show EM1’s malignant cell distribution overlaid on the hematoxylin and eosin (H&E) image. (C) Heatmap of canonical marker expression in EM1 regions, validating transcriptional segregation and matching pathologist-defined regions. Markers of leukemic populations and dermis regions show shared expression profiles. Unsupervised cluster overlap is represented as pie charts, with pathology annotation. (D) Phenotypic staining (Opal multiplex immunofluorescent) on near-adjacent sections validating the spatial distribution of CD33 (malignant cells), CD68, and CD71, which is consistent with the Visium malignant signature (spot-level) and CD68 and CD71 expression patterns. Scale bars: 1 mm (whole-slide panels) and 100 μm (selected region panels).

Journal: iScience

Article Title: Integrative spatial multi-omics reveal niche-specific inflammatory signaling and differentiation hierarchies in AML

doi: 10.1016/j.isci.2025.114289

Figure Lengend Snippet: Spatial multi-omics profiling identifies leukemic infiltration and tissue composition in extramedullary acute myeloid leukemia samples (A) Unsupervised clustering of the extramedullary sample EM1 into 3 spatial clusters (left) compared against the pathology-based annotation (right; indicating a composition of leukemia, dermis, epidermis, and gland). The adjusted rand index (ARI; 0.51) reflects moderate agreement between the clusters and pathology annotations. (B) Spatial deconvolution scores obtained using the SpaCET algorithm show EM1’s malignant cell distribution overlaid on the hematoxylin and eosin (H&E) image. (C) Heatmap of canonical marker expression in EM1 regions, validating transcriptional segregation and matching pathologist-defined regions. Markers of leukemic populations and dermis regions show shared expression profiles. Unsupervised cluster overlap is represented as pie charts, with pathology annotation. (D) Phenotypic staining (Opal multiplex immunofluorescent) on near-adjacent sections validating the spatial distribution of CD33 (malignant cells), CD68, and CD71, which is consistent with the Visium malignant signature (spot-level) and CD68 and CD71 expression patterns. Scale bars: 1 mm (whole-slide panels) and 100 μm (selected region panels).

Article Snippet: Visium Spatial Transcriptomics Data , This paper , GEO: GSE279576.

Techniques: Biomarker Discovery, Marker, Expressing, Staining, Multiplex Assay

Inflammatory microenvironment analysis reveals region-specific signatures in bone marrow and extramedullary tissues from patients with acute myeloid leukemia (A) Distribution of spatial inflammation classes in BM1 and EM1, based on composite inflammation scores from inflammation-related hallmark pathways (Inflammatory response, IL6/JAK/STAT3 signaling, TNF-α/NF-κB signaling, IFN-γ response, IFN-α response, Complement, IL2/STAT5 signaling). Classes were defined using Jenks' natural breaks optimization. (B) Mean activity comparison of individual inflammatory related pathways in spots with high-inflammatory activity revealed the highest activity of IFN-γ response in EM tissue. Complement pathway activity is higher in BM1 when compared with the EM1 inflammatory niche. ∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001, ∗∗∗∗ p < 0.0001, Wilcoxon rank-sum test. (C) Boxplots of inflammation scores across the 3 clusters in BM1 (left) and EM1 (right). Each cluster displays significantly different levels of inflammatory activity; leukemia-enriched cluster 3 in BM1 and cluster 1 in EM1 have higher inflammation scores. ∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001, ∗∗∗∗ p < 0.0001, Wilcoxon rank-sum test. (D) IL-6 staining (Opal multiplex fluorescent immunohistochemistry [mfIHC]) in whole-slide images (left) of BM1 (top) and EM1 (bottom) and corresponding magnified regions (center), aligned with Visium spot-level composite inflammation score (right). Scale bars: 1 mm (whole-slide panels) and 100 μm (selected region panels). (E) Scatterplots showing the correlation of IL-6 protein staining intensity (mfIHC IL-6) with the composite inflammation score in BM1 (left) and EM1 (right). IL-6 levels are higher in high-inflammation regions in both BM1 and EM1. (F) Dot plot showing the localization of T cell subtypes (exhausted, CD8 + dysfunction, senescence, regulatory T cells [Treg]) based on inflammation class in BM1 and EM1.

Journal: iScience

Article Title: Integrative spatial multi-omics reveal niche-specific inflammatory signaling and differentiation hierarchies in AML

doi: 10.1016/j.isci.2025.114289

Figure Lengend Snippet: Inflammatory microenvironment analysis reveals region-specific signatures in bone marrow and extramedullary tissues from patients with acute myeloid leukemia (A) Distribution of spatial inflammation classes in BM1 and EM1, based on composite inflammation scores from inflammation-related hallmark pathways (Inflammatory response, IL6/JAK/STAT3 signaling, TNF-α/NF-κB signaling, IFN-γ response, IFN-α response, Complement, IL2/STAT5 signaling). Classes were defined using Jenks' natural breaks optimization. (B) Mean activity comparison of individual inflammatory related pathways in spots with high-inflammatory activity revealed the highest activity of IFN-γ response in EM tissue. Complement pathway activity is higher in BM1 when compared with the EM1 inflammatory niche. ∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001, ∗∗∗∗ p < 0.0001, Wilcoxon rank-sum test. (C) Boxplots of inflammation scores across the 3 clusters in BM1 (left) and EM1 (right). Each cluster displays significantly different levels of inflammatory activity; leukemia-enriched cluster 3 in BM1 and cluster 1 in EM1 have higher inflammation scores. ∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001, ∗∗∗∗ p < 0.0001, Wilcoxon rank-sum test. (D) IL-6 staining (Opal multiplex fluorescent immunohistochemistry [mfIHC]) in whole-slide images (left) of BM1 (top) and EM1 (bottom) and corresponding magnified regions (center), aligned with Visium spot-level composite inflammation score (right). Scale bars: 1 mm (whole-slide panels) and 100 μm (selected region panels). (E) Scatterplots showing the correlation of IL-6 protein staining intensity (mfIHC IL-6) with the composite inflammation score in BM1 (left) and EM1 (right). IL-6 levels are higher in high-inflammation regions in both BM1 and EM1. (F) Dot plot showing the localization of T cell subtypes (exhausted, CD8 + dysfunction, senescence, regulatory T cells [Treg]) based on inflammation class in BM1 and EM1.

Article Snippet: Visium Spatial Transcriptomics Data , This paper , GEO: GSE279576.

Techniques: Activity Assay, Comparison, Staining, Multiplex Assay, Immunohistochemistry

Chemokine signaling through the CXCL12-CXCR4 axis is linked to inflammatory niches and trans-differentiation in acute myeloid leukemia (A) Spatial and chord diagrams of the strength of interactions among acute myeloid leukemia (AML) cells, granulocyte-monocyte progenitors (GMP), and monocytes through the CXCL12-CXCR4 axis, as predicted by CellChat. (B) Boxplots of the expression levels of CXCL12 and CXCR4 in BM1, stratified by inflammation class (top), and corresponding spot-level expression maps (bottom) for the bone marrow sample BM1. Red spots indicate higher expression levels. (C and D) Whole-slide images of Opal multiplex fluorescent immunohistochemistry (mfIHC; left) for CXCR4 (turquoise) and CXCL12 (magenta) overlaid with DAPI (blue), alongside magnified Opal regions and Visium-based gene expression maps (right) in BM1 (C) and the extramedullary sample EM1 (D). Scale bars: 1 mm (whole-slide panels) and 100 μm (selected region panels). (E) Scatterplot shows the positive correlation of the PI3K/Akt/mTOR pathway score with the combined CXCL12-CXCR4 co-expression score (R = 0.50, p < 2.2e-16). Colors denote inflammation class. (F) Relationship between CXCR4 expression and inflammation score in EM1 (R = 0.19, p < 2.2e-16). Spatial maps show the distribution of CXCR4 expression. (G) Boxplots comparing CXCR4 protein signal intensity (mfIHC) across inflammation classes in BM1 (left) and EM1 (right). Spot-level images illustrate higher CXCR4 signal intensities in high-inflammation areas. (H) Sections 1 and 2 represent adjacent serial sections of the same EM1 biopsy embedded on a single Visium capture area. Visium ST visualization of PI3K/Akt/mTOR pathway (left) and trans -differentiation pathway (right) activity in these EM1 sections, revealing elevated pathway scores in high-inflammation and leukemic regions.

Journal: iScience

Article Title: Integrative spatial multi-omics reveal niche-specific inflammatory signaling and differentiation hierarchies in AML

doi: 10.1016/j.isci.2025.114289

Figure Lengend Snippet: Chemokine signaling through the CXCL12-CXCR4 axis is linked to inflammatory niches and trans-differentiation in acute myeloid leukemia (A) Spatial and chord diagrams of the strength of interactions among acute myeloid leukemia (AML) cells, granulocyte-monocyte progenitors (GMP), and monocytes through the CXCL12-CXCR4 axis, as predicted by CellChat. (B) Boxplots of the expression levels of CXCL12 and CXCR4 in BM1, stratified by inflammation class (top), and corresponding spot-level expression maps (bottom) for the bone marrow sample BM1. Red spots indicate higher expression levels. (C and D) Whole-slide images of Opal multiplex fluorescent immunohistochemistry (mfIHC; left) for CXCR4 (turquoise) and CXCL12 (magenta) overlaid with DAPI (blue), alongside magnified Opal regions and Visium-based gene expression maps (right) in BM1 (C) and the extramedullary sample EM1 (D). Scale bars: 1 mm (whole-slide panels) and 100 μm (selected region panels). (E) Scatterplot shows the positive correlation of the PI3K/Akt/mTOR pathway score with the combined CXCL12-CXCR4 co-expression score (R = 0.50, p < 2.2e-16). Colors denote inflammation class. (F) Relationship between CXCR4 expression and inflammation score in EM1 (R = 0.19, p < 2.2e-16). Spatial maps show the distribution of CXCR4 expression. (G) Boxplots comparing CXCR4 protein signal intensity (mfIHC) across inflammation classes in BM1 (left) and EM1 (right). Spot-level images illustrate higher CXCR4 signal intensities in high-inflammation areas. (H) Sections 1 and 2 represent adjacent serial sections of the same EM1 biopsy embedded on a single Visium capture area. Visium ST visualization of PI3K/Akt/mTOR pathway (left) and trans -differentiation pathway (right) activity in these EM1 sections, revealing elevated pathway scores in high-inflammation and leukemic regions.

Article Snippet: Visium Spatial Transcriptomics Data , This paper , GEO: GSE279576.

Techniques: Expressing, Multiplex Assay, Immunohistochemistry, Gene Expression, Activity Assay

Bone proximity analysis reveals the spatial distribution of acute myeloid leukemia cells in different differentiation states (A) Representative spatial map of SpatialTime calculated distances from trabeculae overlaid with hematoxylin and eosin (H&E) image. (B) Boxplots show deconvolution scores of primitive-like, granulocyte-monocyte progenitor (GMP)-like, and committed-like acute myeloid leukemia (AML) cells relative to their distance from bone in Visium data. ∗ p < 0.05, ∗∗∗∗ p < 0.0001, Wilcoxon rank-sum test. (C) GeoMx analysis of AML deconvolution in bone marrow regions from 3 patients with AML (PT3, PT4, PT5). D, distal (dark red); P, proximal (dark blue); B, bone (white). Stacked bar plots represent cell type deconvolution within distal and proximal regions. Scale bars: 250 μm. (D) Line graphs show proportions of primitive-like and GMP-like cells relative to distance from bone.

Journal: iScience

Article Title: Integrative spatial multi-omics reveal niche-specific inflammatory signaling and differentiation hierarchies in AML

doi: 10.1016/j.isci.2025.114289

Figure Lengend Snippet: Bone proximity analysis reveals the spatial distribution of acute myeloid leukemia cells in different differentiation states (A) Representative spatial map of SpatialTime calculated distances from trabeculae overlaid with hematoxylin and eosin (H&E) image. (B) Boxplots show deconvolution scores of primitive-like, granulocyte-monocyte progenitor (GMP)-like, and committed-like acute myeloid leukemia (AML) cells relative to their distance from bone in Visium data. ∗ p < 0.05, ∗∗∗∗ p < 0.0001, Wilcoxon rank-sum test. (C) GeoMx analysis of AML deconvolution in bone marrow regions from 3 patients with AML (PT3, PT4, PT5). D, distal (dark red); P, proximal (dark blue); B, bone (white). Stacked bar plots represent cell type deconvolution within distal and proximal regions. Scale bars: 250 μm. (D) Line graphs show proportions of primitive-like and GMP-like cells relative to distance from bone.

Article Snippet: Visium Spatial Transcriptomics Data , This paper , GEO: GSE279576.

Techniques:

(a) Overview of the clustering approach and characterization of identified regions. i) Features are extracted from each tiles of H&E slides using the histology foundation model H-optimus-0. ii) Tile features from each slide are clustered using K-means, trained on the discovery cohort (cohort A) and applied to the validation cohorts (cohorts B and C). iii) Expert neuropathologists review and annotate each cluster to define distinct regions. (b) Methods overview. Left : Graphical representation of the methods used to align paired H&E and Visium spatial transcriptomics for cell type annotation in 31 samples from Cohort B from the MOSAIC consortium. Right: Overview of the method used to categorize tiles according to their tile score and tile prediction in 3 classes: high score + high prediction = Long Survival, high score + low prediction = Short Survival, all others = Non-informative. (c) Schematic representation of common spatial arrangements of the regions within tumor sections. (d) Proportion of tiles associated with short survival, long survival, and non-informative categories in the validation cohort in the different regions (excluding the 4 regions with necrosis, hemorrhage, and artefacts) (cohorts B and C). (e) Distribution of regions among Long Survival and Short Survival tiles in the validation cohorts (cohorts B and C). (f) Average of cell count per cell type in the different tissue regions as estimated using the output weights of the cell2loc deconvolution algorithm and 31 samples from cohort B with paired H&E, Visium spatial transcriptomics, and scRNAseq. Rare cell types (median counts < 0.1) were excluded.

Journal: bioRxiv

Article Title: Histology and spatial transcriptomic integration revealed infiltration zone with specific cell composition as a prognostic hotspot in glioblastoma

doi: 10.1101/2025.10.08.681087

Figure Lengend Snippet: (a) Overview of the clustering approach and characterization of identified regions. i) Features are extracted from each tiles of H&E slides using the histology foundation model H-optimus-0. ii) Tile features from each slide are clustered using K-means, trained on the discovery cohort (cohort A) and applied to the validation cohorts (cohorts B and C). iii) Expert neuropathologists review and annotate each cluster to define distinct regions. (b) Methods overview. Left : Graphical representation of the methods used to align paired H&E and Visium spatial transcriptomics for cell type annotation in 31 samples from Cohort B from the MOSAIC consortium. Right: Overview of the method used to categorize tiles according to their tile score and tile prediction in 3 classes: high score + high prediction = Long Survival, high score + low prediction = Short Survival, all others = Non-informative. (c) Schematic representation of common spatial arrangements of the regions within tumor sections. (d) Proportion of tiles associated with short survival, long survival, and non-informative categories in the validation cohort in the different regions (excluding the 4 regions with necrosis, hemorrhage, and artefacts) (cohorts B and C). (e) Distribution of regions among Long Survival and Short Survival tiles in the validation cohorts (cohorts B and C). (f) Average of cell count per cell type in the different tissue regions as estimated using the output weights of the cell2loc deconvolution algorithm and 31 samples from cohort B with paired H&E, Visium spatial transcriptomics, and scRNAseq. Rare cell types (median counts < 0.1) were excluded.

Article Snippet: Differential expression analysis (DEA) was performed on 10x Visium spatial transcriptomics data using a 2-step approach.

Techniques: Biomarker Discovery, Cell Counting

(a) Distributions of the number of nuclei per tile and the median nuclear area per tile across 457 samples in cohort A (training dataset). Only tiles with a tile score > 0 were included. Purple lines denote thresholds optimized on the training set; arrows indicate the direction associated with shorter survival. *** indicates p value < 0.001 for a Mann Whitney U test between two boxplots. (b) Proportion of biomarker-positive tiles defined as infiltrated white matter (liWM) regions with >30 nuclei and a median nuclear area <40 µm² among all tumor tiles in short- and long-surviving patients from cohorts B and C (testing dataset). (c) Representative H&E slide and corresponding Visium spatial transcriptomics overlay (validation dataset) showing biomarker positive liWM tiles (dark purple) in comparison to adjacent liWM tissue (purple). (d) Cell type composition per tile in biomarker-positive liWM tiles versus non informative liWM tiles in the n=31 patients of the MOSAIC dataset. Cell types with median count per tile <0.1 were excluded for clarity. (e) Pathways enriched in short survival liWM vs non informative liWM. Raster plot of GSEA showing leading genes of each pathways by log2FoldChange. Only pathways with FDR<0.2 were included (See GSEA results and all genes in Supplementary table I and Table J [Supplementary material]). (f) Volcano plot for DEA of short survival liWM vs non informative liWM.

Journal: bioRxiv

Article Title: Histology and spatial transcriptomic integration revealed infiltration zone with specific cell composition as a prognostic hotspot in glioblastoma

doi: 10.1101/2025.10.08.681087

Figure Lengend Snippet: (a) Distributions of the number of nuclei per tile and the median nuclear area per tile across 457 samples in cohort A (training dataset). Only tiles with a tile score > 0 were included. Purple lines denote thresholds optimized on the training set; arrows indicate the direction associated with shorter survival. *** indicates p value < 0.001 for a Mann Whitney U test between two boxplots. (b) Proportion of biomarker-positive tiles defined as infiltrated white matter (liWM) regions with >30 nuclei and a median nuclear area <40 µm² among all tumor tiles in short- and long-surviving patients from cohorts B and C (testing dataset). (c) Representative H&E slide and corresponding Visium spatial transcriptomics overlay (validation dataset) showing biomarker positive liWM tiles (dark purple) in comparison to adjacent liWM tissue (purple). (d) Cell type composition per tile in biomarker-positive liWM tiles versus non informative liWM tiles in the n=31 patients of the MOSAIC dataset. Cell types with median count per tile <0.1 were excluded for clarity. (e) Pathways enriched in short survival liWM vs non informative liWM. Raster plot of GSEA showing leading genes of each pathways by log2FoldChange. Only pathways with FDR<0.2 were included (See GSEA results and all genes in Supplementary table I and Table J [Supplementary material]). (f) Volcano plot for DEA of short survival liWM vs non informative liWM.

Article Snippet: Differential expression analysis (DEA) was performed on 10x Visium spatial transcriptomics data using a 2-step approach.

Techniques: MANN-WHITNEY, Biomarker Discovery, Comparison

(a). In cohort A, proportion of tiles associated with short survival, long survival, and non-informative categories in pooled tissue subtypes Tumor 1.0, Tumor 1.1, and Tumor 2.1. (b). Within pooled tissue subtypes Tumor 1.0, Tumor 1.1, and Tumor 2.1, distributions of nuclear morphology features like nuclear density (number of nuclei per tile) and nuclear size (median nuclear area in µm²) and median nuclear circularity. (c) Representative images of the annotated vessels subtypes patterns on H&E slides (20X): (i) thin endothelium capillary (TEC), (ii) hyperplasic endothelium capillary (HEC), (iii) Microvascular Proliferation (MVP) (iv) Thin Endothelium Wide Lumen (TEWL) (v) Hyperplasic Endothelium Wide Lumen (HEWL) (d) Relative ratios of tiles with vessels or specific vessels subtypes within tumor regions (1.0, 1.1 and 2.1) for each patients across the two survival group (2y < OS and OS > 3y). Distributions are computed for the training set (cohort A) and validation set (cohort B&C). Vessels presence and vessels subtypes detection are obtained after inference by the vessels detection and vessels subtypes classifier models on the entire cohorts. (e) Distribution of cell count per cell type in the tumor regions (1.0, 1.1 and 2.1) for short-survival, long-survival and non-informative tiles. Distributions of cell counts statistically significantly different (Mann–Whitney test; threshold at 0.05) are indicated with a star. 31 samples from Cohort B with paired H&E, Visium spatial transcriptomics, and scRNAseq were used. Cell types with low range of mean count variations across tissue regions were removed for clarity. (f) Illustration of the pathway enrichment analyses of the differential expressed genes between long- and short-survival. Negative fold change is associated with worse survival outcomes and positive with better survival outcomes. (See GSEA results and all genes in supplementary table L [Supplementary material])

Journal: bioRxiv

Article Title: Histology and spatial transcriptomic integration revealed infiltration zone with specific cell composition as a prognostic hotspot in glioblastoma

doi: 10.1101/2025.10.08.681087

Figure Lengend Snippet: (a). In cohort A, proportion of tiles associated with short survival, long survival, and non-informative categories in pooled tissue subtypes Tumor 1.0, Tumor 1.1, and Tumor 2.1. (b). Within pooled tissue subtypes Tumor 1.0, Tumor 1.1, and Tumor 2.1, distributions of nuclear morphology features like nuclear density (number of nuclei per tile) and nuclear size (median nuclear area in µm²) and median nuclear circularity. (c) Representative images of the annotated vessels subtypes patterns on H&E slides (20X): (i) thin endothelium capillary (TEC), (ii) hyperplasic endothelium capillary (HEC), (iii) Microvascular Proliferation (MVP) (iv) Thin Endothelium Wide Lumen (TEWL) (v) Hyperplasic Endothelium Wide Lumen (HEWL) (d) Relative ratios of tiles with vessels or specific vessels subtypes within tumor regions (1.0, 1.1 and 2.1) for each patients across the two survival group (2y < OS and OS > 3y). Distributions are computed for the training set (cohort A) and validation set (cohort B&C). Vessels presence and vessels subtypes detection are obtained after inference by the vessels detection and vessels subtypes classifier models on the entire cohorts. (e) Distribution of cell count per cell type in the tumor regions (1.0, 1.1 and 2.1) for short-survival, long-survival and non-informative tiles. Distributions of cell counts statistically significantly different (Mann–Whitney test; threshold at 0.05) are indicated with a star. 31 samples from Cohort B with paired H&E, Visium spatial transcriptomics, and scRNAseq were used. Cell types with low range of mean count variations across tissue regions were removed for clarity. (f) Illustration of the pathway enrichment analyses of the differential expressed genes between long- and short-survival. Negative fold change is associated with worse survival outcomes and positive with better survival outcomes. (See GSEA results and all genes in supplementary table L [Supplementary material])

Article Snippet: Differential expression analysis (DEA) was performed on 10x Visium spatial transcriptomics data using a 2-step approach.

Techniques: Biomarker Discovery, Cell Counting, MANN-WHITNEY

(A–C) UMAP plots of Visium spatial spots from SCC samples. (A) Spots colored by sample origin. (B) Spots labeled by spatial zones (core, edge, surrounding). (C) Grouped zones highlight transcriptomic patterns at tumor boundaries. (D) Heatmap showing spatial proximity among tumor zones. Higher intensity indicates closer spatial interaction, illustrating SCC microenvironment organization. (E) Ligand–receptor analysis identifies fibroblast-derived SFRP2 interacting with LRP5/6, suggesting Wnt signaling activation. (F) Spatial maps of SFRP2⁺ fibroblasts and Wnt signaling scores in SCC. Enrichment in tumor cores and interfaces points to fibroblast-driven modulation of the tumor niche.

Journal: bioRxiv

Article Title: Integrated Single-Cell and Spatial Transcriptomic Analysis Reveals an Aging-Associated Fibroblast Subtype Linked to Tumor Progression in Human Skin

doi: 10.1101/2025.07.09.663827

Figure Lengend Snippet: (A–C) UMAP plots of Visium spatial spots from SCC samples. (A) Spots colored by sample origin. (B) Spots labeled by spatial zones (core, edge, surrounding). (C) Grouped zones highlight transcriptomic patterns at tumor boundaries. (D) Heatmap showing spatial proximity among tumor zones. Higher intensity indicates closer spatial interaction, illustrating SCC microenvironment organization. (E) Ligand–receptor analysis identifies fibroblast-derived SFRP2 interacting with LRP5/6, suggesting Wnt signaling activation. (F) Spatial maps of SFRP2⁺ fibroblasts and Wnt signaling scores in SCC. Enrichment in tumor cores and interfaces points to fibroblast-driven modulation of the tumor niche.

Article Snippet: Raw Visium spatial transcriptomics data (BCL and FASTQ files) were processed using the SpaceRanger v1.3.1 pipeline (10x Genomics).

Techniques: Labeling, Derivative Assay, Activation Assay

EAC primary and metastatic samples show a diverse landscape of TME and malignant cells in transcriptomic and epigenetic data (A) Schematic representation of the study workflow. Biopsies from 10 patients in our discovery cohort, including normal adjacent tissue (NAT), primary tissue, and metastatic samples, were subjected to single-nuclei RNA and ATAC sequencing using 10X Chromium technology. For a subset of these patients as well as three additional patients, matched primary and metastatic samples were profiled with 10X Visium and 10X Xenium spatial transcriptomics (ST) technologies. For single-nuclei data, cells were annotated by cell type and categorized into malignant and TME components. TME subtypes were linked to metastasis, with validation against an external pan-cancer fibroblast atlas. The malignant cell components underwent analysis using consensus non-negative matrix factorization (cNMF) to uncover malignant programs, which were further characterized for transcriptional and epigenetic heterogeneity at a single-cell and spatial level and candidate master transcription factors. External validation was performed in two single-cell validation cohorts, , and associations with clinical and molecular characteristics, as well as survival, were assessed in three bulk validation cohorts. , , (B) Uniform manifold approximation and projection (UMAP) representation of the full cohort in Harmony-corrected integrated transcriptomic data, with major cell type compartments labeled and cell counts indicated. (C) Proportion of major cell types in each sample based on transcriptomic data, with percentages for compartments representing over 5% of the total sample composition. (D) UMAP representation of the full cohort in Harmony-corrected integrated ATAC data, with cell type annotations transferred from the RNA annotations. “NA” denotes cells without paired associated RNA information. (E) Proportion of major cell types in each sample based on ATAC data, with percentages for compartments representing over 5% of the total sample composition.

Journal: Cell Reports Medicine

Article Title: Cell states and neighborhoods in distinct clinical stages of primary and metastatic esophageal adenocarcinoma

doi: 10.1016/j.xcrm.2025.102188

Figure Lengend Snippet: EAC primary and metastatic samples show a diverse landscape of TME and malignant cells in transcriptomic and epigenetic data (A) Schematic representation of the study workflow. Biopsies from 10 patients in our discovery cohort, including normal adjacent tissue (NAT), primary tissue, and metastatic samples, were subjected to single-nuclei RNA and ATAC sequencing using 10X Chromium technology. For a subset of these patients as well as three additional patients, matched primary and metastatic samples were profiled with 10X Visium and 10X Xenium spatial transcriptomics (ST) technologies. For single-nuclei data, cells were annotated by cell type and categorized into malignant and TME components. TME subtypes were linked to metastasis, with validation against an external pan-cancer fibroblast atlas. The malignant cell components underwent analysis using consensus non-negative matrix factorization (cNMF) to uncover malignant programs, which were further characterized for transcriptional and epigenetic heterogeneity at a single-cell and spatial level and candidate master transcription factors. External validation was performed in two single-cell validation cohorts, , and associations with clinical and molecular characteristics, as well as survival, were assessed in three bulk validation cohorts. , , (B) Uniform manifold approximation and projection (UMAP) representation of the full cohort in Harmony-corrected integrated transcriptomic data, with major cell type compartments labeled and cell counts indicated. (C) Proportion of major cell types in each sample based on transcriptomic data, with percentages for compartments representing over 5% of the total sample composition. (D) UMAP representation of the full cohort in Harmony-corrected integrated ATAC data, with cell type annotations transferred from the RNA annotations. “NA” denotes cells without paired associated RNA information. (E) Proportion of major cell types in each sample based on ATAC data, with percentages for compartments representing over 5% of the total sample composition.

Article Snippet: Raw and processed spatial transcriptomics Visium data , This paper , Zenodo: 10.5281/zenodo.15341263.

Techniques: Sequencing, Biomarker Discovery, Labeling

Journal: Cell Reports Medicine

Article Title: Cell states and neighborhoods in distinct clinical stages of primary and metastatic esophageal adenocarcinoma

doi: 10.1016/j.xcrm.2025.102188

Figure Lengend Snippet: Single-nuclei-derived transcriptional programs highlight different spatial regions of EAC tumors (A) Spatial transcriptomics (ST) slides of P8 primary tumor A, colored according to cNMF program score and the CNV-derived label. For each spot, we infer the CNV profile with inferCNV and assign spots to tumor, mixed, and normal status. cNMF scores are computed as the average Z score of signature genes using the deconvolved carcinoma-specific gene expression profile of spots derived with Cell2Location. (B) Average cNMF score according to the position of the spots compared to the tumor-leading edge. For each tumor spot, we compute the distance to the edge as the shortest path to a normal or mixed spot. The distribution of cNMF scores with standard error is represented for normal spots, mixed spots, and spots of a certain distance to the edge. (C) cNMF scores for carcinoma cells and cell type annotations in a subset of Xenium-profiled samples (P4_A, P11_B, and P12_D). cNMF scores are computed as the average Z score across all carcinoma cells of signature genes included in the Xenium panel. (D) CellCharter cluster assignments for a subset of Xenium-profiled samples (P4_A, P11_B, and P12_D). (E) CellCharter cluster cell type proportion: each cell is assigned a cluster and we represent the proportion of the different cell types in each of the CellCharter clusters. (F) Distribution of cNMF scores of carcinoma cells belonging to the CellCharter clusters with a substantial amount of carcinoma cells (>5%), CC1, CC2, CC3, and CC4. For all comparisons, Mann-Whitney U p < 0.000005.

Article Snippet: Raw and processed spatial transcriptomics Visium data , This paper , Zenodo: 10.5281/zenodo.15341263.

Techniques: Derivative Assay, Gene Expression, MANN-WHITNEY

Journal: Scientific Reports

Article Title: Image guided construction of a common coordinate framework for spatial transcriptome data

doi: 10.1038/s41598-025-01862-x

Figure Lengend Snippet: Schema representation of STaCker. The workflow ( a ) takes as inputs the tissue images from a pair of reference and moving spatial transcriptome slices, combined with the contour maps generated upon the gene expression profiles from the respective slices. The resulting composite images are subsequently aligned through a deep neural network-based registration module. The registration module outputs the inferred deformation field to align the spatial coordinates of the spots/cells in the moving slice. The architecture of the registration module ( b ) takes a four-level contracting path and a four-level expanding path with skip connections at all levels. The final layer of the decoder is further convoluted to generate the spatial velocity field followed by a vector integration to output the deformation field for the alignment. Synthetic images with segmentation label maps (Methods) are used to train the module. The moved label map, after applying the deformation field to the moving label map, is compared to the reference label map. The difference constitutes the key component in the loss function.

Article Snippet: The Visium spatial transcriptome data of a human lymph node sample were obtained from 10 × Genomics website [ https://cf.10xgenomics.com/samples/spatial-exp/1.1.0/V1_Human_Lymph_Node/V1_Human_Lymph_Node_filtered_feature_bc_matrix.h5 ; https://cf.10xgenomics.com/samples/spatial-exp/1.1.0/V1_Human_Lymph_Node/V1_Human_Lymph_Node_spatial.tar.gz ].

Techniques: Generated, Gene Expression, Plasmid Preparation

Journal: Scientific Reports

Article Title: Image guided construction of a common coordinate framework for spatial transcriptome data

doi: 10.1038/s41598-025-01862-x

Figure Lengend Snippet: Evaluation of STaCker in aligning digitally warped spatial transcriptome slices of mouse brain. ( a ) The reference is a mouse sagittal posterior brain slice profiled by 10 × Genomics Visium platform. It was digitally warped using Simplex noises to a low (noise amplitude = 5, NCC of the deformed image = 0.61), medium (noise amplitude = 10, NCC of the deformed image = 0.57), or high (noise amplitude = 15, NCC of the deformed image = 0.54 level to generate a series of moving slices (noise frequency remains 1 for all warping). ( b ) The discordance between the spatial coordinates of the spots in each moving slice and those in the reference slice is quantified by the MSE (Methods) shown in the bar plots. STaCker, together with previously published methods STUtility, PASTE, and GPSA, was applied to align each of the moving spatial transcriptome slices to the reference. The spots’ coordinates before (blue crosses) or after the alignment (red crosses) are displayed together with the reference spot coordinates (gray dots) to aid the visual comparison. The post-alignment MSEs from each method are illustrated in the bar plots. Value from STaCker is the mean over five runs, shown together with the standard errors as error bars. STaCker’s MSE is significantly lower than that of all other programs (one sample t-test p -values < = 1e-3).

Techniques: Slice Preparation, Comparison

Journal: Scientific Reports

Article Title: Image guided construction of a common coordinate framework for spatial transcriptome data

doi: 10.1038/s41598-025-01862-x

Figure Lengend Snippet: Performance of STaCker in the de novo alignment of spatial transcriptome slices. The top row displays four moving spatial transcriptome slices that were independently warped from a reference slice taken from the mouse posterior brain used in Fig. , with the spot coordinates shown as crosses over the tissue images (slice 1: red, slice 2: green, slice 3: blue, slice 4: orange). The warping was conducted using random-seeded Simplex noises with an amplitude of 15 and a frequency of 1. The mean pairwise NCCs among the tissue images of the moving slices is 0.198. The average pairwise MSE among the spot coordinates in the moving slices is 0.10. The bottom row illustrates the spot coordinates from four slices before the alignment (“Unaligned coordinates”) and after the alignment by STaCker, STUtility, PASTE, GPSA, respectively, using the same colors and cross symbols as shown in the top row. The post-alignment average MSE over all six pairs of slices is 0.043, 0.119, 0.098, 0.601 for STaCker, STUtility, PASTE, and GPSA, respectively.

Techniques:

Journal: Scientific Reports

Article Title: Image guided construction of a common coordinate framework for spatial transcriptome data

doi: 10.1038/s41598-025-01862-x

Figure Lengend Snippet: Coordinate consolidation of real spatial transcriptome slices from a human dorsolateral prefrontal cortex (DLPFC). ( a ): Four serial dissections of the dorsolateral prefrontal cortex. ( b ): Superimposed spatial coordinates of the QC-validated tissue spots from four DLPFC slices before and after alignment by different methods, color-coded by their tissue domain annotations in the original publication. ( c - d ): Quantitative evaluation of tissue domain consistency across the four slices before and after alignment, using Spatial Coherence Score ( c ) and Mean Pairwise Adjusted Rand Index ( d ). STUtility does not offer de novo alignments so their values are averaged over four alignments, each using a different slice as the fixed template, with error bars marking the standard errors. ( e ): Spatial patterns of representative genes before and after alignment by STaCker and other programs. The displayed expression values are the natural logarithm transformation of the normalized UMI count (10 4 total UMI counts per spot). ( f ): Comparison of Moran’s I spatial autocorrelations of representative genes before and after alignment by various programs.

Techniques: Expressing, Transformation Assay, Comparison

Journal: Scientific Reports

Article Title: Image guided construction of a common coordinate framework for spatial transcriptome data

doi: 10.1038/s41598-025-01862-x

Figure Lengend Snippet: Coordinate consolidation of real spatial transcriptome slices from independent replicates of mouse olfactory bulbs. ( a ): Four biological replicates of dissected mouse olfactory bulbs. ( b ): Superimposed spatial coordinates of the QC-validated tissue spots from four slices before and after alignment by different methods, color-coded by the tissue domain annotations derived upon the transcriptome of the spots. ( c - d ): Quantitative evaluation of tissue domain consistency across the four slices before and after alignment, using Spatial Coherence Score ( c ) and Mean Pairwise Adjusted Rand Index ( d ). STUtility does not offer de novo alignments so their values are averaged over four alignments, each using a different slice as the fixed template. The standard errors are shown as error bars. ( e ): Spatial patterns of representative genes before and after alignment by STaCker and the other programs. The displayed expression is after the natural logarithm transformation of the normalized UMI count (10 4 total UMI counts per spot). ( f ): Comparison of Moran’s I spatial autocorrelations of representative genes before and after alignment by various programs.

Techniques: Derivative Assay, Expressing, Transformation Assay, Comparison

Journal: Scientific Reports

Article Title: Image guided construction of a common coordinate framework for spatial transcriptome data

doi: 10.1038/s41598-025-01862-x

Figure Lengend Snippet: Alignment of ISH-based spatial transcriptome slices. ( a ) MERFISH slices from three mouse brain samples, illustrated using the DAPI staining images. ( b ) Superimposed spatial coordinates of cells in the three slices before and after alignment by different methods. Cells are color-coded by their niches, defined based on the gene profiling of the cells (see Methods). ( c ) Quantitative evaluation of tissue domain consistency across the three slices before and after alignment, using Spatial Coherence Score and Mean Pairwise Adjusted Rand Index. STUtility and STalign do not offer de novo alignments so their values are averaged over three alignments each with a different slice as the fixed template, with standard errors shown as error bars. ( d ) Spatial patterns of four representative genes before and after alignment by STaCker, STUtility and STalign. The displayed expression is after the natural logarithm transformation of the normalized UMI count (10 4 total UMI counts per cell). ( e ) Comparison of Moran’s I autocorrelation (left panel) and Gene Coherence Score (right panel) of the four representative genes from the slices aligned by STaCker, STUtility or STalign. Values from STUtility and STalign are the average over three alignments each using a different slice as the fixed template, with standard errors shown as error bars. ( f ) Boxplots of the Moran’s I autocorrelation (left panel) and Gene Coherence Score (right panel) of all non-randomly distributed genes (Moran’s I p -value < = 0.01) over the slices aligned by STaCker, STUtility or STalign. The top and bottom edges of the box represent the 3rd and 1st quantiles, with the horizontal line inside denoting the median. The ends of the whisker mark the 1.5 times interquartile range, calculated as the difference between the 3rd and 1st quartiles, from the box edges. Data points beyond the whisker range are represented as dots. STalign was executed using the same parameters applied to the same MERFISH dataset in the original publication. ( g ) Comparison of the spatial coherence in gene expressions after the alignment by STaCker and STalign. Genes that show significantly higher Moran’s I correlation (left panel) or Gene Coherence Score (GCS, right panel) after alignment with STaCker are marked with orange dots, while genes with significantly elevated values for these metrics following STalign alignment are indicated by blue dots. The dashed line represents the significance cutoff (0.05) for the two-sided Student’s t-test p -value.

Techniques: Staining, Expressing, Transformation Assay, Comparison, Whisker Assay

Journal: Scientific Reports

Article Title: Image guided construction of a common coordinate framework for spatial transcriptome data

doi: 10.1038/s41598-025-01862-x

Figure Lengend Snippet: Alignment across different spatial transcriptome platforms. ( a ) Two mouse brain hemispheres profiled using 10 × Genomics Visium and Xenium, shown as the acquired H&E and DAPI images, respectively. ( b ) Superimposed spatial coordinates of spots (red circles, Visium slice) or cells (blue crosses, Xenium slice) before and after alignment by different methods. ( c ) Spatial patterns of the representative genes before and after alignment by STaCker, STUtility or STalign. The positions of the spots in the Visium slice (red) and the 55-micron × 55-micron pseudo-spots in the Xenium slice (blue) are displayed. At each spot or pseudo-spot, the expression of a gene is divided by the maximum expression of that gene on the slice, converting it to a value within 0 and 1. The scaled gene expressions are comparable across platforms and thus used for visualization and quantitative evaluation. ( d ) Moran’s I autocorrelation (upper panel) and the Gene Coherence Score (lower panel) of the representative genes from the slices aligned by STaCker, STUtility or STalign. Values for STUtility and STalign that do not offer de novo alignment are the average over alignments each using one of the slices as the reference, with the standard errors shown as error bars. ( e ) Boxplots of the Moran’s I autocorrelation score (upper panel) and the Gene Coherence Score (lower panel) of all non-randomly distributed genes (Moran’s I p -value < = 0.01) over the slices aligned by STaCker, STUtility or STalign. For both metrics, the mean of the distribution in STaCker is significantly higher than that in STalign (two-sided student t-test p -value < = 6e-5) and in STUtility (two-sided student t-test p -value < = 2e-3). In all boxplots, the top and bottom edges of the box represent the 3rd and 1st quantiles with the horizontal line inside to denote the median. The whiskers extend to 1.5 times the interquartile range (IQR), which is the difference between the 3rd and 1st quantiles, from the box edges. Data points outside the whisker range are displayed as dots. ( f ) Comparison of spatial coherence in gene expressions following alignment using STaCker and STalign. Genes exhibiting significantly higher Moran’s I correlation (left panel) or Gene Coherence Score (right panel) after alignment with STaCker are depicted with orange dots. Genes with significantly increased values for the two metrics after alignment with STalign are shown with blue dots. The dashed line indicates the significance cutoff (0.05) for the two-sided Student’s t-test p -value.

Techniques: Expressing, Whisker Assay, Comparison

Review

Structured Review

Images

1) Product Images from "Image guided construction of a common coordinate framework for spatial transcriptome data"

Similar Products

Image Search Results