Identifying Spatial Co-occurrence in Healthy and InflAmed tissues (ISCHIA)

Abstract

Spatial transcriptomics (ST) techniques are able to chart the distribution and localization of cell types and RNA molecules across a tissue. Sequencing-based methods allow unbiased capturing of RNA molecules at barcoded spots. While the coarse resolution of these techniquee is considered a disadvantage, we argue that the inherent proximity of transcriptomes captured on spots can be leveraged to reconstruct cellular networks. To this end, we developed ISCHIA (Identifying Spatial Co-occurrence in Healthy and InflAmed tissues), a computational framework to analyze the spatial co-occurrence of cell types and transcript species in the tissue environment. Co-occurrence analysis is complementary to differential gene expression, as it does not depend on the abundance of a given cell type or the transcript expression levels, but rather on their spatial arrangement in the tissue. We applied ISCHIA to analyze co-occurrence of cell types, ligands and receptors in a sequencing-based ST dataset of human ulcerative colitis, and validated our findings on matched hybridization-based data. We uncover inflammation-induced cellular networks involving M-cell and fibroblasts, as well as ligand-receptor interactions enriched in the inflamed human colon, and their associated gene signatures. Our results highlight the hypothesis-generating power and broad applicability of co-occurrence analysis on spatial transcriptomics data.

In vivosingle-cell CRISPR uncovers distinct TNF-α programs in clonal expansion and tumorigenesis

“…Conversely, Mmp10-negative TSKs localized to differentiated regions at the surface of the tumor, distant from the invasive front (Figure 4n). Co-occurrence analysis 35 confirmed a cellular neighborhood of Mmp10-positive TSKs with fibroblasts and endothelial cells, in line with a fibrovascular niche surrounding TSK cells (Figure 4m), aligning with a previous report 31 . Of note, Mmp10-positive TSKs correlated with G2M clusters, while Mmp10-negative TSKs showed a negative correlation with G2M, suggesting that Mmp10-positive TSKs reside in proliferative tumor areas (Figure 4m, Extended Data Figure 7d).…”

“…Seurat R package (version 4.9.9.9041) was used to plot these cell-type deconvolution results. Spatial co-occurrence of cell-types was analyzed with the ISCHIA package 35 (version 1.0.0.0), and resulting probabilities for positive (Pgt) or negative (Plt) co-occurrence were plotted in -log10 scale. Bioinformatic processing of the scRNA-seq data The raw sequencing data comprising the BCL files were demultiplexed using the Illumina bcl2fastq program v2.20 with default options allowing one mismatch of the sample barcode sequences.…”

Identification of ligand and receptor interactions in CKD and MASH through the integration of single cell and spatial transcriptomics

“…We performed co-occurrence analyses to investigate non-random associations between cell types within the tissue. To this end, we utilized the ISCHIA method, a computational combinatorics framework that assigns a quantitative property to the interaction potential of cell types by computing their spatial co-occurrence [ 36 ]. This method enables us to identify pairs of cell types that exhibit positive co-occurrence in specific cellular neighborhoods, indicating that the observed co-occurrence is higher than expected by chance.…”

The inflammatory micro-environment induced by targeted CNS radiotherapy is underpinned by disruption of DNA methylation

“…Therefore, we next used the computational tool STdeconvolve to determine the cell-type composition of these clusters. This method can effectively recover cell-type transcriptional profiles and their proportional representation within ST spots without reliance on external single-cell transcriptomics references [26][27][28]. For each deconvolved cell type, annotation was performed using a combination of cluster markers and most highly expressed genes, compared with expression in known cell types using the Human Protein Atlas and STAB database [29], as well as the spatial location of cell types, confirmed by a neuropathologist (Fig 3 C, D and supplementary fig.…”