Journal: Nucleic Acids Research

Article Title: STAMarker: determining spatial domain-specific variable genes with saliency maps in deep learning

doi: 10.1093/nar/gkad801

Figure Lengend Snippet: STAMarker reveals the domain-specific SVGs on the mouse olfactory bulb dataset. ( A ) Spatial domains identified by STAMarker. The laminar organization of the mouse olfactory bulb is clearly shown. The identified spatial domains were annotated by the Allen Reference Atlas. ( B ) Histogram of the number of spatial domains to which the SVGs identified by STAMarker belong. ( C ) UpSet plot of the numbers of SVGs identified by STAMarker and six compared methods. STAMarker identified 311 SVGs. ( D ) Heatmap showing the eight modules that were clustered by the 67 domain-specific SVGs. ( E ) Visualization of the domain-specific gene modules by the first principal component of the saliency maps. ( F ) Visualization of the representative spatial domain-specific SVGs.

Article Snippet: The human lymph node and mouse olfactory bulb datasets are available at 10x genomics website 10xgenomics.com/resources/datasets hippocampus dataset of the J20 mouse model generated by Slide-seq V2 is accessible at https://singlecell.broadinstitute.org/single_cell/study/SCP1663/cell-type-specific-inference-of-differential-expression-in-spatial-transcriptomics .

Techniques:

Journal: Small Methods

Article Title: Robust Spatial Cell‐Type Deconvolution with Qualitative Reference for Spatial Transcriptomics

doi: 10.1002/smtd.202401145

Figure Lengend Snippet: Analysis of mouse olfactory bulb data. a) H&E staining of the olfactory bulb (top) and the deconvolution results of all candidate methods displayed by the spatial scatter pie plot of cell‐type composition on each spatial location. The examined cell types were granule cells (GC), olfactory sensory neurons (OSNs), periglomerular cells (PGC), mitral/tufted cells (M‐TC), and external plexiform layer interneurons (EPL‐IN) b) Manual annotation of anatomic layers (top), including the granule cell layer (GCL), the mitral cell layer (MCL), the glomerular layer (GL), and the nerve layer (ONL), and the spatial domains of different deconvolution methods visualized by spatial scatters of specific domain types. c) Performance comparison between candidate deconvolution methods, including QR‐SIDE, STdeconvolve, CARDfree, RCTD, CARD, SPOTlight, and spatialDWLS in terms of NMI (left) and ARI (right). d) UMAP plots of gene expression for Topic 1, 2, 3 identified by QR‐SIDE. The color scheme of each topic domain was the same as in (b). e) The heatmap of normalized expression level for the top 10 DE genes for topic domain 1, 2, 3. f) The correlation between DE genes of identified domains and marker genes of each cell type. g) The mean expression level of an example marker gene list for QR‐SIDE, where Tyro3 was included as the interference marker gene of cell type GC. h) Left and middle panels: The estimated spot‐separable η scores of correct marker Penk and misclassified marker Tyro3 . Right panel: The line plots of mean η of all markers across all spatial spots and the RMSE between estimated cell‐type composition by varying the inclusion of top 3‐7 marker genes of each cell type as the input gene list and the deconvolution results using high‐quality marker genes (as shown in a).

Article Snippet: These include the mouse olfactory bulb ST data from Spatial Transcriptomics v1.0 ( https://www.spatialresearch.org ), the four human hepatocellular carcinoma Visium datasets ( https://www.ncbi.nlm.nih.gov/sra?linkname=bioproject_sra_all&from_uid=858545 ), mouse anterior brain 10x Visium data ( https://support.10xgenomics.com/spatial‐gene‐expression/datasets/1.0.0/V1_Mouse_Brain_Sagittal_Anterior ), and mouse posterior brain 10x Visium data ( https://support.10xgenomics.com/spatial‐gene‐expression/datasets/1.0.0/V1_Mouse_Brain_Sagittal_Posterior ).

Techniques: Staining, Comparison, Gene Expression, Expressing, Marker

Journal: Bioinformatics

Article Title: JUMP: replicability analysis of high-throughput experiments with applications to spatial transcriptomic studies

doi: 10.1093/bioinformatics/btad366

Figure Lengend Snippet: Analysis and validation results of the mouse olfactory bulb data measured with ST technology and 10× Visium technology. (a) Venn diagram shows the number of replicable SVGs identified by different methods at FDR level 0.05 and the intersection of discoveries. (b) Three distinct spatial expression patterns based on the 807 replicable SVGs identified by JUMP in the ST study (top) and 10× Visium study (bottom). Each pattern summarizes the relative expression levels across spatial spots. The corresponding hematoxylin and eosin staining images for the two studies are shown in the right panel. (c) Spatial expression patterns of three representative genes identified by JUMP, corresponding to Patterns I–III, respectively. In situ hybridization images of corresponding genes obtained from the Allen Brain Atlas ( atlas.brain-map.org ) are shown in the top panel. (d) The box plot shows Moran’s I statistic of the 189 replicable SVGs additionally identified by JUMP and that of all genes based on the ST study (left) and the 10× Visium study (right). (e) The bar chart displays the number of replicable SVGs additionally identified by JUMP and BH compared to that identified by all three methods. They were validated in two reference gene lists from the Harmonizome database: one from the Allen Brain Atlas dataset and the other from the BioGPS dataset. (f) The bubble plot shows the GO enrichment analysis result of JUMP, including different GO term categories: BP, CC and MF. The horizontal dashed line represents the FDR level 0.01. The size of a bubble represents the counts of corresponding gene sets.

Article Snippet: The mouse olfactory bulb 10× Visium dataset is available in the 10× Visium spatial gene expression repository at https://www.10xgenomics.com/resources/datasets/adult-mouse-olfactory-bulb-1-standard-1 .

Techniques: Biomarker Discovery, Expressing, Staining, In Situ Hybridization