Journal: Bioinformatics Advances

Article Title: Revealing tissue architecture through the hypercomplex Fourier analysis of spatial transcriptomics data

doi: 10.1093/bioadv/vbaf191

Figure Lengend Snippet: Visualization of 10× Visium HD spatial transcriptomics data for coronal mouse brain slice. The H&E stained image is displayed behind a visualization of the Visium spots colored according to the number of UMI counts.

Article Snippet: To illustrate the proposed quaternion model and the ST visualization, rotation and hypercomplex Fourier analysis applications, we analysed a 10× Visium HD ST dataset generated on a mouse coronal brain slice (results for a mouse brain sagittal Visium dataset, a mouse kidney Visium dataset, and human ovarian cancer Vizgen MERSCOPE dataset are included as vingettes with the QSC R package).

Techniques: Slice Preparation, Staining

Journal: Bioinformatics Advances

Article Title: Revealing tissue architecture through the hypercomplex Fourier analysis of spatial transcriptomics data

doi: 10.1093/bioadv/vbaf191

Figure Lengend Snippet: Visualization of quaternion mapping for coronal mouse brain slice Visium HD data. The left panel shows the SVD model and the right panel shows a mapping based on the rank 20 reconstructed expression of the GFAP, Reln, and Neurod6 genes, which are markers for astrocytes, interneurons and excitatory neurons and are mapped to red, green, and blue respectively.

Article Snippet: To illustrate the proposed quaternion model and the ST visualization, rotation and hypercomplex Fourier analysis applications, we analysed a 10× Visium HD ST dataset generated on a mouse coronal brain slice (results for a mouse brain sagittal Visium dataset, a mouse kidney Visium dataset, and human ovarian cancer Vizgen MERSCOPE dataset are included as vingettes with the QSC R package).

Techniques: Slice Preparation, Expressing

Journal: Nucleic Acids Research

Article Title: Benchmarking sketching methods on spatial transcriptomics data

doi: 10.1093/nar/gkag434

Figure Lengend Snippet: Overall sketching performance for 0.10 sampling fraction across datasets. ( A ) Spatial scatter plots of real datasets colored by cell type or cluster label. ( B ) Heatmap of rank-sums for each method aggregated by metric across all real world datasets. Low rank indicates best performance for that metric. ( C ) Spatial scatter plots of simulated Visium HD-like and Xenium like datasets. ( D ) Heatmap of rank-sums for each method aggregated by metric across all simulated datasets. Low rank indicates best performance for that metric.

Article Snippet: Mouse embryo: We downloaded the whole mouse embryo Visium HD dataset from the publicly available datasets on the 10x website https://www.10xgenomics.com/datasets/visium-hd-three-prime-mouse-embryo-fresh-frozen (last accessed date: 12 February 2025).

Techniques: Sampling

Journal: Nucleic Acids Research

Article Title: Benchmarking sketching methods on spatial transcriptomics data

doi: 10.1093/nar/gkag434

Figure Lengend Snippet: Retained cell type/cluster label proportions at 0.10 sketching fraction for ( A ) Merfish mouse ovary; ( B ) Merfish sagittal mouse brain; ( C ) Xenium human breast cancer; ( D ) Xenium human lung; ( E ) Xenium whole mouse pup; ( F ) Visium HD coronal mouse brain; ( G ) Visium HD mouse embryo; ( H ) Visium HD ovarian cancer.

Article Snippet: Mouse embryo: We downloaded the whole mouse embryo Visium HD dataset from the publicly available datasets on the 10x website https://www.10xgenomics.com/datasets/visium-hd-three-prime-mouse-embryo-fresh-frozen (last accessed date: 12 February 2025).

Techniques:

Journal: Nucleic Acids Research

Article Title: Benchmarking sketching methods on spatial transcriptomics data

doi: 10.1093/nar/gkag434

Figure Lengend Snippet: Quantification of transcriptomic and coordinate Hausdorff distance at 0.10 sampling fraction for real datasets. ( A ) Quantification of imaging based (Merfish, Xenium) dataset’s Hausdorff distances. ( B ) Quntification of sequencing/spot based (Visium HD) dataset’s Hausdorff distances. Each boxplot represents one sketching method, with individual points corresponding to results from 10 independent runs with different random seeds.

Article Snippet: Mouse embryo: We downloaded the whole mouse embryo Visium HD dataset from the publicly available datasets on the 10x website https://www.10xgenomics.com/datasets/visium-hd-three-prime-mouse-embryo-fresh-frozen (last accessed date: 12 February 2025).

Techniques: Sampling, Imaging, Sequencing

Journal: Nature Communications

Article Title: Reconstructing single-cell resolution from spatial transcriptomics with CellRefiner

doi: 10.1038/s41467-026-70090-2

Figure Lengend Snippet: a Visium ST data from a slice of the mouse cortex. b The distribution of annotated cell types from the scRNA-seq data on the CellRefiner output. c Segmentation of CellRefiner reconstruction of mouse cortex Visium data, using SpaceFlow, with colors corresponding to clusters. d Ripley’s L for spatial organization on Visium and CellRefiner. e Neighborhood enrichment score on spatial proximity of clusters for CellRefiner output. f ST data from murine lymph node using Visium. Source data are provided as a Source Data file.

Article Snippet: The mouse kidney Visium data is available at the 10X Genomics website ( https://www.10xgenomics.com/datasets/mouse-kidney-section-coronal-1-standard-1-1-0 ).

Techniques:

Journal: Nature Communications

Article Title: Reconstructing single-cell resolution from spatial transcriptomics with CellRefiner

doi: 10.1038/s41467-026-70090-2

Figure Lengend Snippet: a The single-cell resolution spatial map of cells reconstructed by CellRefiner using a paired Visium data and scRNA-seq data. The analysis rediscovered several contact-based signaling activities confirmed by prior knowledge, including EPHB, NOTCH, ICAM, and CDH. b CellRefiner also identified several highly active junction-related signaling including EPHA, MPZ, CD39, CD46, DESMOSOME, and JAM. Source data are provided as a Source Data file.

Article Snippet: The mouse kidney Visium data is available at the 10X Genomics website ( https://www.10xgenomics.com/datasets/mouse-kidney-section-coronal-1-standard-1-1-0 ).

Techniques: Single Cell

Journal: Cell reports methods

Article Title: Precise detection of cell-type-specific domains in spatial transcriptomics.

doi: 10.1016/j.crmeth.2024.100841

Figure Lengend Snippet: Figure 2. De-spot identified similar cell-type-specific domains in multiple mouse brain slices (A) Free frozen tissue (slice 1) and FFPE tissue (slice 2). Annotations of anatomic mouse brain structure are from the Allen Brain Atlas. (B) Previously published single-cell profiles, including seven annotated cell types: astrocytes-ependymal, endothelial-mural, interneurons, microglia, oligo- dendrocytes, pyramidal CA1, and pyramidal SS. (C) 3D Landscapes of slice 1 and slice 2 generated by Giotto. The color of each domain corresponds to the cell types with the same color in (B). (D) 3D Landscapes of slice 1 and slice 2 generated by Seurat. (E) 3D Landscapes of slice 1 and slice 2 generated by CARD.

Article Snippet: REAGENT or RESOURCE SOURCE IDENTIFIER Deposited data scRNA-seq data of mouse brain cortex Zeisel et al.39 GEO: GSE60361 10x Visium data of mouse brain cortex (free frozen) 10x Genomics https://www.10xgenomics.com/resources/ datasets/mouse-brain-section-coronal-1standard-1-0-0 10x Visium data of mouse brain cortex (FFPE) 10x Genomics https://www.10xgenomics.com/resources/ datasets/adult-mouse-brain-ffpe-1standard-1-3-0 scRNA-seq data of mouse kidney Wu et al.47 GEO: GSE119531 SRT data of mouse kidney (FFPE) 10x Genomics https://www.10xgenomics.com/resources/ datasets/adult-mouse-kidney-ffpe-1standard-1-3-0 ST and paired scRNA-seq data of PDAC Moncada et al.30 GEO: GSE111672 MERFISH data of mouse brain cortex Zhuang et al.68 https://cellxgene.cziscience.com/ collections/31937775-06024e52-a799b6acdd2bac2e Stereo-seq data of mouse brain cortex STOmicsDB70 https://db.cngb.org/stomics/datasets/ STDS0000234 10x Visium data of human breast cancer Wu et al.58 GEO: GSE176078 Software and algorithms R (v4.1.3) R Core Team https://www.r-project.org/ Python (v3.9.7) Python Software Foundation https://www.python.org/ CARD (v1.0) Ma et al.31 https://github.com/YMa-lab/CARD Seurat (v4.3.0) Butler et al.24 https://satijalab.org/seurat Giotto (v1.1.0) Dries et al.25 https://github.com/RubD/Giotto/ Squidpy (v1.2.2) Palla et al.26 https://squidpy.readthedocs.io/en/stable/ Scanpy (v1.9.1) Wolf et al.27 https://scanpy.readthedocs.io/en/stable/ Cell2Location (v0.1) Kleshchevnikov et al.16 https://github.com/BayraktarLab/ cell2location Stlearn (v0.4.6) Pham et al.14 https://github.com/ BiomedicalMachineLearning/stLearn SpaGCN (v1.2.2) Hu et al.12 https://github.com/jianhuupenn/SpaGCN BayesSpace (v1.4.1) Zhao et al.11 https://github.com/edward130603/ BayesSpace SEDR (v1.0.0) Fu et al.13 https://github.com/JinmiaoChenLab/SEDR MENDER (v1.1) Yuan25 https://github.com/yuanzhiyuan/MENDER BASS (v1.1.0.016) Li and Zhou24 https://github.com/zhengli09/BASS SPOTlight (v0.99.8) Elosua-Bayes et al.19 https://github.com/MarcElosua/SPOTlight StereoScope (v0.3) Andersson et al.18 https://github.com/almaan/stereoscope RCTD (v2.0.0) Cable et al.17 https://github.com/dmcable/spacexr SPROD (v1.0) Wang et al.34 https://github.com/yunguan-wang/SPROD SPCS (inline) Liu et al.36 https://github.com/Usos/SPCS SpotClean (v0.99.2) Ni et al.33 https://github.com/zijianni/SpotClean SCDD (v1.0.0) Liu et al.35 https://github.com/lyotvincent/SCDD Matplotlib (v3.5.1) Matplotlib development team https://matplotlib.org De-spot (v1.0.0) This paper https://zenodo.org/doi/10.5281/zenodo.

Techniques: Generated

Journal: Cell reports methods

Article Title: Precise detection of cell-type-specific domains in spatial transcriptomics.

doi: 10.1016/j.crmeth.2024.100841

Figure Lengend Snippet: Figure 3. De-spot detected co-localized cell-type-specific domains using multiple profiles (A) FFPE tissue of the mouse kidney. The slice is histologically annotated to the renal cortex, renal medulla 1, renal medulla 2, and pelvis. (B) Previously published mouse kidney profiles generated by scRNA-seq and small nuclear RNA sequencing, containing 13 cell types annotated by Wu et al.47 CD-PC, collecting duct principal cells; CNT, connecting tubule; DCT, distal convoluted tubules; EC, endothelial cells; IC, intercalated cells; LHAL, the loop of Henle ascending loop; LHDL, the loop of Henle descending loop; MC, mesangial cells; MØ, macrophages; PT, proximal tubules; Pod, po- docytes.

Article Snippet: REAGENT or RESOURCE SOURCE IDENTIFIER Deposited data scRNA-seq data of mouse brain cortex Zeisel et al.39 GEO: GSE60361 10x Visium data of mouse brain cortex (free frozen) 10x Genomics https://www.10xgenomics.com/resources/ datasets/mouse-brain-section-coronal-1standard-1-0-0 10x Visium data of mouse brain cortex (FFPE) 10x Genomics https://www.10xgenomics.com/resources/ datasets/adult-mouse-brain-ffpe-1standard-1-3-0 scRNA-seq data of mouse kidney Wu et al.47 GEO: GSE119531 SRT data of mouse kidney (FFPE) 10x Genomics https://www.10xgenomics.com/resources/ datasets/adult-mouse-kidney-ffpe-1standard-1-3-0 ST and paired scRNA-seq data of PDAC Moncada et al.30 GEO: GSE111672 MERFISH data of mouse brain cortex Zhuang et al.68 https://cellxgene.cziscience.com/ collections/31937775-06024e52-a799b6acdd2bac2e Stereo-seq data of mouse brain cortex STOmicsDB70 https://db.cngb.org/stomics/datasets/ STDS0000234 10x Visium data of human breast cancer Wu et al.58 GEO: GSE176078 Software and algorithms R (v4.1.3) R Core Team https://www.r-project.org/ Python (v3.9.7) Python Software Foundation https://www.python.org/ CARD (v1.0) Ma et al.31 https://github.com/YMa-lab/CARD Seurat (v4.3.0) Butler et al.24 https://satijalab.org/seurat Giotto (v1.1.0) Dries et al.25 https://github.com/RubD/Giotto/ Squidpy (v1.2.2) Palla et al.26 https://squidpy.readthedocs.io/en/stable/ Scanpy (v1.9.1) Wolf et al.27 https://scanpy.readthedocs.io/en/stable/ Cell2Location (v0.1) Kleshchevnikov et al.16 https://github.com/BayraktarLab/ cell2location Stlearn (v0.4.6) Pham et al.14 https://github.com/ BiomedicalMachineLearning/stLearn SpaGCN (v1.2.2) Hu et al.12 https://github.com/jianhuupenn/SpaGCN BayesSpace (v1.4.1) Zhao et al.11 https://github.com/edward130603/ BayesSpace SEDR (v1.0.0) Fu et al.13 https://github.com/JinmiaoChenLab/SEDR MENDER (v1.1) Yuan25 https://github.com/yuanzhiyuan/MENDER BASS (v1.1.0.016) Li and Zhou24 https://github.com/zhengli09/BASS SPOTlight (v0.99.8) Elosua-Bayes et al.19 https://github.com/MarcElosua/SPOTlight StereoScope (v0.3) Andersson et al.18 https://github.com/almaan/stereoscope RCTD (v2.0.0) Cable et al.17 https://github.com/dmcable/spacexr SPROD (v1.0) Wang et al.34 https://github.com/yunguan-wang/SPROD SPCS (inline) Liu et al.36 https://github.com/Usos/SPCS SpotClean (v0.99.2) Ni et al.33 https://github.com/zijianni/SpotClean SCDD (v1.0.0) Liu et al.35 https://github.com/lyotvincent/SCDD Matplotlib (v3.5.1) Matplotlib development team https://matplotlib.org De-spot (v1.0.0) This paper https://zenodo.org/doi/10.5281/zenodo.

Techniques: Generated, RNA Sequencing

Journal: bioRxiv

Article Title: FlashDeconv enables atlas-scale, multi-resolution spatial deconvolution via structure-preserving sketching

doi: 10.64898/2025.12.22.696108

Figure Lengend Snippet: (a) Abundance invariance test. Ranking stability of oligodendrocyte markers (top 20 genes by expression in oligodendrocytes) in mouse brain scRNA-seq (31,053 genes) as cell population is downsampled from 26.7% to 0.4%. Rank denotes average position when all genes are sorted by score (rank 1 = highest). Variance-based ranking (blue) degrades from rank 115 to 240 as abundance decreases—a two-fold deterioration. Leverage-score ranking (red) remains stable at rank ~ 150 regardless of population size, demonstrating true decoupling of biological identity from numerical prevalence. (b) The variance-leverage plane. Classification of 31,053 genes by variance (x-axis) and leverage score (y-axis). Four quadrants emerge: structurally informative “GOLD” genes (green, low variance/high leverage) include vascular markers ( Cldn5, Rgs5, Ly6a, Abcb1a, Hspb1 ) that define rare anatomical structures; variance-dominated “NOISE” genes (red, high variance/low leverage) contain 35% unannotated Gm -series transcripts compared to only 6% in the GOLD set, indicating that high variance alone does not ensure cell-type discriminative power. (c) Functional enrichment analysis. GO Biological Process enrichment reveals GOLD genes are significantly enriched for regulation of angiogenesis (FDR-adjusted p = 2.8 × 10 −6 ), endothelial cell differentiation (FDR-adjusted p = 2.1 × 10 −4 ), vasculogenesis, and blood vessel morphogenesis. NOISE genes show zero significant GO terms at FDR-adjusted p < 0.05. Genome-wide cell type specificity analysis further confirms that GOLD genes systematically target rare populations (median 0.27% abundance) versus NOISE genes (0.51%; p = 3.25 × 10 −25 ), with Endothelial cells as the top target—validating leverage as an unsupervised metric for biological distinctiveness. (d) Spatial verification on Visium. Top row: GOLD genes ( Cldn5, Ly6a, Rgs5 ) reconstruct clear vascular anatomical structures on mouse brain Visium sections (spatial structure score = 1.33). Bottom row: NOISE genes exhibit random, speckle-like distribution patterns (structure score = 0.87; Mann-Whitney p = 5.6 × 10 −5 ). This visual contrast demonstrates that leverage selects for genuine biological structure rather than technical variation.

Article Snippet: The Visium HD Mouse Small Intestine (FFPE) dataset was obtained from 10x Genomics ( https://www.10xgenomics.com/datasets/visium-hd-cytassist-gene-expression-libraries-of-mouse-intestine ).

Techniques: Expressing, Functional Assay, Cell Differentiation, Genome Wide, MANN-WHITNEY

Journal: bioRxiv

Article Title: FlashDeconv enables atlas-scale, multi-resolution spatial deconvolution via structure-preserving sketching

doi: 10.64898/2025.12.22.696108

Figure Lengend Snippet: (a) FlashDeconv performance across resolutions. Left: Number of measurement units at each bin size. Middle: Processing time demonstrating scalability to 350,000+ spots. Right: Signal purity (fraction of spots with > 80% single cell type) collapses from 61.5% at 8 µ m to 13.3% at 16 µ m. (b) Spatial maps of enterocyte proportions at 8, 16, and 32 µ m resolution on mouse small intestine. Fine anatomical detail visible at 8 µ m becomes progressively blurred at coarser resolutions. (c) Resolution sensitivity varies by cell type. Stem cells (red) show the steepest decline in spatial coherence (Moran’s I), while Paneth cells (blue) retain spatial structure. Shaded region indicates the 8–16 µ m transition zone. (d) Cryptvillus boundary validation. Gradient sharpness decreases by 77% from 8 µ m to 16 µ m, quantifying anatomical blurring. (e) Spatial binning induces spurious colocalization. Paneth and Goblet cells show weak mutual exclusion at 8 µ m ( r = −0.12, p < 10 −100 ) but appear strongly colocalized at 64 µ m ( r = +0.80, p < 10 −100 )—a correlation sign flip that could lead to incorrect biological conclusions about cell-cell interactions. Data: Visium HD Mouse Small Intestine (10x Genomics), scRNA-seq reference from Haber et al. 2017.

Article Snippet: The Visium HD Mouse Small Intestine (FFPE) dataset was obtained from 10x Genomics ( https://www.10xgenomics.com/datasets/visium-hd-cytassist-gene-expression-libraries-of-mouse-intestine ).

Techniques: Biomarker Discovery

Journal: bioRxiv

Article Title: FlashDeconv enables atlas-scale, multi-resolution spatial deconvolution via structure-preserving sketching

doi: 10.64898/2025.12.22.696108

Figure Lengend Snippet: (a) HVG blindness ranking across intestinal cell types. HVG blindness is defined as the difference in mean percentile rank of a cell type’s marker genes under variance-based versus leverage-based selection; positive values indicate systematic underweighting by HVG. Tuft (brush) cells exhibit the highest HVG blindness (21 percentile points). (b) Spatial distribution of Tuft cells at 8 µ m resolution reveals focal niches (red spots) with proportions up to 61%. (c) Stem cell distribution at 8 µ m shows concentration at crypt bases. (d) Resolution sensitivity of Tuft cell detection. Maximum proportion decreases from 61% (8 µ m) to 4% (128 µ m), rendering focal niches undetectable at conventional resolution. (e) Colocalization analysis reveals Tuft cell hotspots are enriched 16.8-fold for stem cells and 15.3-fold for enteroendocrine cells ( p < 10 −4 , permutation test), but depleted for differentiated cell types (enterocytes 0.11×, goblet cells 0.10×). (f) Spatial zoom showing Tuft-Stem co-localization at crypt bases (blue: Stem-high, pink: co-localized). Tuft cells rarely appear without adjacent stem cells, consistent with their intimate niche association. Data: Visium HD Mouse Small Intestine (10x Genomics).

Article Snippet: The Visium HD Mouse Small Intestine (FFPE) dataset was obtained from 10x Genomics ( https://www.10xgenomics.com/datasets/visium-hd-cytassist-gene-expression-libraries-of-mouse-intestine ).

Techniques: Marker, Selection, Concentration Assay

Journal: iMeta

Article Title: A comprehensive benchmarking for spatially resolved transcriptomics clustering methods across variable technologies, organs, and replicates

doi: 10.1002/imt2.70084

Figure Lengend Snippet: Overview of datasets and methods for benchmarking. (A) Workflow of the benchmarking study. Real datasets collected from public databases, together with simulated datasets generated with SRTsim and scCube, were used to evaluate spatial clustering methods across multiple application scenarios. Fourteen spatial clustering methods, spanning both probabilistic statistics and neural network‐based methods, were compared for accuracy across technologies, organs, biological replicates, and simulated spatial patterns. (B) Summary of real datasets used in the study. Datasets obtained from ST, 10× Visium, Slide‐seq, Stereo‐seq, Visium HD, seqFISH+, STARmap, MERFISH, CosMx, and Xenium technologies are shown, including the number of slices, spatial resolution, number of spots, genes, and sparsity per slice. Bar lengths represent the mean number of spots, and error bars indicate standard deviation.

Article Snippet: The adult mouse kidney (FFPE) 10× Visium data set was downloaded from https://www.10xgenomics.com/resources/datasets/adult-mouse-kidney-ffpe-1-standard-1-3-0 , containing 1 tissue slice.

Techniques: Generated, Standard Deviation

Journal: iMeta

Article Title: A comprehensive benchmarking for spatially resolved transcriptomics clustering methods across variable technologies, organs, and replicates

doi: 10.1002/imt2.70084

Figure Lengend Snippet: Performance comparison with SRT datasets across organs. (A) Ground‐truth annotations for representative slices from each organ in the 10× Visium datasets. (B) Ground‐truth annotations and clustering results from all methods on slice 151676 of the DLPFC 10× Visium data set. (C) Box plots compare methods on all 10× Visium datasets of variable organs with Adjusted Rand Index (ARI) and Normalized Mutual Information (NMI). Centerline: median; box limits: upper and lower quartiles; whiskers: 1.5× interquartile range. Results of kidney and skin datasets are presented in Figure S16 due to lower confidence in the ground truth. (D) Ground‐truth annotations for representative slices from each organ in the Slide‐seq datasets. (E) Ground truth and clustering result from each method on a representative Hippocampus Slide‐seq slice. (F) Comparison of methods across all Slide‐seq datasets from different organs, with clustering accuracy measured by ARI and NMI. Centerline: median; box limits: upper and lower quartiles; whiskers: 1.5× interquartile range. SRT, spatially resolved transcriptomics.

Article Snippet: The adult mouse kidney (FFPE) 10× Visium data set was downloaded from https://www.10xgenomics.com/resources/datasets/adult-mouse-kidney-ffpe-1-standard-1-3-0 , containing 1 tissue slice.

Techniques: Comparison

Journal: iMeta

Article Title: A comprehensive benchmarking for spatially resolved transcriptomics clustering methods across variable technologies, organs, and replicates

doi: 10.1002/imt2.70084

Figure Lengend Snippet: Performance comparison with SRT datasets across biological replicates. (A) Simulation of variable replicates based on DLPFC 10× Visium datasets. Box plot shows Normalized Mutual Information (NMI) scores on simulated datasets. (B) Simulation of variable replicates based on the hypothalamus MERFISH data set. Box plot shows NMI scores on simulated datasets. (C) Design of simulation for neighborhood‐changing replicates by merging adjacent clusters. Box plot shows NMI scores on simulated datasets. (D) Design of simulation for neighborhood‐changing replicates by adding new clusters. Box plot shows NMI scores on simulated datasets. The box represents the interquartile range, the horizontal line inside the box indicates the median, and the whiskers extend to 1.5× interquartile range. SRT, spatially resolved transcriptomics.

Article Snippet: The adult mouse kidney (FFPE) 10× Visium data set was downloaded from https://www.10xgenomics.com/resources/datasets/adult-mouse-kidney-ffpe-1-standard-1-3-0 , containing 1 tissue slice.

Techniques: Comparison