Journal: Cell Genomics

Article Title: MultiSP deciphers tissue structure and multicellular communication from spatial multi-omics data

doi: 10.1016/j.xgen.2026.101141

Figure Lengend Snippet:

Article Snippet: 10× Visium human tonsil gene and protein expression data , 10× Genomics , https://www.10xgenomics.com/resources/datasets/gene-protein-expression-library-of-human-tonsil-cytassist-ffpe-2-standard.

Techniques: Expressing, Software

Journal: bioRxiv

Article Title: Single-cell spatial multiomics identifies POSTN + CAFs mediating chemoradiotherapy resistance in rectal cancer

doi: 10.64898/2026.04.30.721803

Figure Lengend Snippet: a, Schematic illustration of the neoadjuvant chemoradiotherapy (CRT) schedule and timing of tumor sampling in the study cohort (N = 24). The CRT regimen consisted of capecitabine in combination with 50.4 Gy of radiation delivered in 28 fractions. “Pre” refers to the time point prior to CRT, “JustAfter” indicates within 2 weeks after the completion of CRT, and “Resection” represents three months post-CRT. Biopsy samples were collected from all 24 patients at the Pre time point and from 4 patients at the JustAfter time point, and surgical samples were obtained from all 24 patients at the Resection time point. Among the 24 patients, 8 achieved a pathological complete response (pCR), 8 achieved a major pathological response (MPR), and 8 were classified as no response (NR). b, Each sample obtained at the designated time points was serially sectioned into four parts (Sections 1–4). Section 1 was used for Xenium analysis, PhenoCycler analysis, and hematoxylin and eosin (H&E) staining. Section 2 was used for Xenium analysis and H&E staining. Section 3 was used for Visium HD spatial transcriptomics on adjacent serial sections. Section 4 was used for whole-exome sequencing. c, Summary matrix showing patient characteristics (age, sex, recurrence status) and the availability of each modality. The purple circles represent samples analyzed at both the Pre and Resection time points, whereas the dark circles indicate samples analyzed at all three time points: Pre, JustAfter, and Resection.

Article Snippet: In addition, we generated transcriptome-wide spatial gene expression profiles using Visium HD (10x Genomics) on adjacent serial sections as a complementary dataset for future integrative analyses; Visium HD data were not used for the primary analyses presented in this study.

Techniques: Sampling, Staining, Spatial Transcriptomics, Sequencing

Journal: bioRxiv

Article Title: Probabilistic coupling of cellular and microenvironmental heterogeneity by masked self-supervised learning

doi: 10.64898/2026.04.21.719876

Figure Lengend Snippet: (a) Spatial visualization comparing k-means clustering of niche representations from methods (Mievformer, NicheCompass, CellCharter, Banksy, GraphST, STAGATE, ENVI, Spatial, PCA). cell type annotations (left). (b) Heatmap showing the Pearson correlation between eight evaluation metrics (CAS, SCAS, NEA, DREC, NCB-GCS, NCB-CAS, NCB-NASW, NCB-MLAMI) and ground-truth Niche ARI across varying simulation parameters. (c) Horizontal bar graphs showing dual recovery (DREC) metrics across methods and five datasets: Xenium GBM, Xenium Lung, Xenium Pancreas, Mouse Brain Stereo-seq, and Mouse Brain Visium HD.

Article Snippet: The Visium HD Mouse Brain data were obtained from the 10x Genomics public dataset repository ( https://www.10xgenomics.com/datasets ).

Techniques:

Journal: Nature Communications

Article Title: FineST: contrastive learning integrates histology and spatial transcriptomics for nuclei-resolved ligand-receptor analysis

doi: 10.1038/s41467-026-70528-7

Figure Lengend Snippet: A HE-stained image of the Visium tissue section and adjacent Xenium section, alongside their alignment. One repeat is performed for the publicly available data. B Pearson correlation between FineST and iStar for all input genes, calculated after aggregating super-resolution data to spot resolution. C Spatial expression plots for OPRPN from left to right: Visium, Xenium, FineST and iStar. FineST enhances the signal relative to Visium and yields results more comparable to Xenium. D Pearson correlation for OPRPN in FineST, corresponding to panel ( C ). Each dot represents a Visium spot ( n = 4992) or overlapping Xenium pseudo-spot ( n = 3958). E Ground-truth cell type annotations at spot (Visium) and single-cell (Xenium) resolution, as reported previously . F FineST's predicted cell types at single-nucleus and sub-spot levels. G FineST accurately identifies the DCIS 2 cell type in a triple-positive receptor ROI. H Pearson correlation for cell type abundance and mean gene expression across each cell type, comparing FineST and iStar. Each dot represents a cell type (Left, n = 19) or gene (Right, n = 65), lines connect matched pairs. Diamond indicates the mean. Statistical significance was assessed by a paired two-sided t -test. I Three marked regions (ROI 1, ROI 2 and ROI 3) dominated by DCIS 1, DCIS 2 and Invasive tumor cells. J Cell type deconvolution from FineST, compared with Xenium ground truth, demonstrates FineST's results visually match the ground truth and outperform Visium’s lower resolution (see Supplementary Fig. ). Alongside, single-cell resolved CCC patterns identified using SparseAEH (cluster number 2) and pathway enrichment analysis for Pattern 0 correspond to interesting cell distributions. K Venn plot of significant LR pairs (FDR < 0.05) interacting in >25% (ROI 1, 5589 cells) or >20% (ROI 2, 3330 cells; ROI 3, 5853 cells) of cells. In total, 103, 146 and 159 pairs were selected for spatial clustering analysis in the three ROIs, respectively. L Comparative analysis of region- and cell-specific LR pairs reveals two unique pairs specific to DCIS 2 and Invasive tumor cells. For panels ( B ), ( H – J ), source data are provided in the Source Data file. Scare bars, 1 mm.

Article Snippet: The raw and processed Visium spatial sequencing data of human hepatocellular carcinoma (HCC) tissues were downloaded from Mendeley Data under accession number skrx2fz79n [ https://data.mendeley.com/datasets/skrx2fz79n/1 ], while the cell type annotations and high-resolution HE-stained images were provided by the original corresponding author.

Techniques: Staining, Expressing, Single Cell, Gene Expression

Journal: Nature Communications

Article Title: FineST: contrastive learning integrates histology and spatial transcriptomics for nuclei-resolved ligand-receptor analysis

doi: 10.1038/s41467-026-70528-7

Figure Lengend Snippet: A , B Spatial plot of seven cell types estimated by cell2location (spot resolution) and FineST (single-cell resolution). A' , B' Zoon-in views of A and B . B'' Zoon-in view of the region marked in B' , where FineST increases resolution from 8 spots to 869 single cells. C Cell type composition from reference scRNA-seq and deconvolution at Visium spot and FineST single-cell levels. D Distribution and PCC of cell type proportions ( n = 7) for Visium spot (PCC = 0.24) and FineST single-cell (PCC = 0.64) resolutions vs reference scRNA-seq. E Pathologists identified 36 spots (of 1331) co-localized with tertiary lymphoid structure (TLS), important for antigen presentation and T cell activation. FineST's single-cell views validate TLS by T and B cell co-localization (see B'' ). F Three example spatially co-expressed LR pairs detected at FineST's single-cell resolution. G Visualization of PVR - TIGIT interaction among local single cells ( z -score FDR < 0.05). H Clustering of 633 significant LR pairs, from single-cell resolution, within selected ROI into three spatial patterns using SpatialDE. I Spatial co-localization of tumor and Treg cells, estimated by cell2location . J Spatially co-expressed LR pairs detected at FineST's single-cell resolution. Scatter plot of global Moran’s R and one-sided z -score p -values (orange: significant, FDR < 0.05, Benjamini-Hochberg correction), with examples highlighted. K Communication strength of CD70 - CD27 at single-cell resolution (color: \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$1-{p}_{{{{\rm{local}}}}{z}_{p}}$$\end{document} 1 − p local z p ; mean strength: 0.45, interacting cells: 5961, occupancy: 49.2%). L Detection of local CCC patterns by clustering significant LR pairs. Top: original spot resolution (332 pairs, SpatialDE); Middle: sub-spot resolution (957 pairs, SparseAEH) and Bottom: single-nucleus resolution (931 pairs, SparseAEH). M Dot plots of enriched pathways in Pattern 0 and Pattern 1 (from L , Bottom), 2-by-2 contingency tables for MHC-I and WNT. Statistical significance was assessed using a one-sided Fisher’s exact test; dot size indicates p -value. N Sankey plot of selected L-R-TF-TG communication pathways. For panels (D , K ), box plots show median (center), IQR (box), whiskers at 1.5 × IQR, and points for seven cell types; violin plots show density, median (white line), and IQR (thick bar). For panels ( A ), ( J , L ), source data are provided in the Source Data file.

Article Snippet: The raw and processed Visium spatial sequencing data of human hepatocellular carcinoma (HCC) tissues were downloaded from Mendeley Data under accession number skrx2fz79n [ https://data.mendeley.com/datasets/skrx2fz79n/1 ], while the cell type annotations and high-resolution HE-stained images were provided by the original corresponding author.

Techniques: Single Cell, Immunopeptidomics, Activation Assay

Journal: Nature Communications

Article Title: FineST: contrastive learning integrates histology and spatial transcriptomics for nuclei-resolved ligand-receptor analysis

doi: 10.1038/s41467-026-70528-7

Figure Lengend Snippet: A , B HE staining and cell type annotation of spatial transcriptomic spots in tumor tissues from an ICB non-responder (P1_T) and responder (P7_T) from the original study . In the non-responder (P1_T), a tumor immune barrier (TIB) structure is formed by SPP1 + macrophages and cancer-associated fibroblasts (CAFs). C , D Spatial signature score of SPP1 + macrophages and CAFs, spatial gene expression of CCR1 , and Pearson correlation between SPP1 + macrophage score and CCR1 expression across all spots (Visium vs FineST). The red line represents the fitted linear regression, and the shaded area corresponds to the 95% confidence interval. Statistical significance was assessed using a two-sided Pearson correlation test. E Scatterplot of global Moran’s R and one-sided z -score p -values for spatially co-expressed LR pairs detected using FineST-enhanced gene expression at spot-level resolution in ROI. Significant pairs (orange) were identified using FDR < 0.05 (Benjamini-Hochberg correction). The CD274-PDCD1 interaction was detected in P1_T only. F Spatial gene expression of ligand CD274 and receptor PDCD1 across all spots (Visium vs FineST) for P1_T (non-responder) and P7_T (responder), respectively. For panels ( A , B , E ), source data are provided in the Source Data file. Part of the panels ( A , B ) is created in BioRender. Huang, Y. (2026) https://BioRender.com/fv0byvk .

Article Snippet: The raw and processed Visium spatial sequencing data of human hepatocellular carcinoma (HCC) tissues were downloaded from Mendeley Data under accession number skrx2fz79n [ https://data.mendeley.com/datasets/skrx2fz79n/1 ], while the cell type annotations and high-resolution HE-stained images were provided by the original corresponding author.

Techniques: Staining, Gene Expression, Expressing