Multimodal Hypersampling of the adult human heart

Analysis code for the manuscript "Multimodal Hypersampling of the adult human heart resolves tissue-specific cell states and disease-associated variants."

We profiled 28 anatomical regions of the adult human heart from 36 donors, combining paired single-nucleus RNA-seq and ATAC-seq (10x Multiome) with spatial transcriptomics (Visium, Visium HD, Xenium) and imaging. This repository holds the custom code referenced in the paper's Code Availability statement, organised by modality.

How to read this repository

This is a curated record of analyses carried out over several years by several people on several compute systems. It is not a single runnable pipeline:

• Notebooks and scripts are not end-to-end runnable. They document the analysis steps and the exact parameters used.
• All file paths are placeholders (/path/to/...). Supply your own paths to the atlas objects, references and intermediate files.
• Inputs are the processed objects released via the Heart Cell Atlas (above), not re-derived here from raw reads.
• Each modality directory has its own README.md with the execution order, inputs, outputs and software for that analysis.

Repository layout

GEX/         single-nucleus RNA-seq: demultiplexing, QC, integration, annotation,
             and downstream analyses (CellPhoneDB, Hotspot, Milo, NOTCH, NOVA1, RUNX1)
ATAC/        single-nucleus ATAC-seq: ArchR processing, chromVAR, SCENIC+/SNP2CELL,
             allele-specific accessibility, and ChromBPNet / AlphaGenome sequence models
VisiumSD/    Visium (55 µm) spatial transcriptomics: niches, deconvolution, OrganAxis
VisiumHD/    Visium HD (2 µm): bin2cell segmentation, TACCO annotation, transmural NMF
Xenium/      Xenium in situ: 5K cell-state transfer and image co-registration
Microscopy/  post-Xenium immunofluorescence: valve macrophage quantification

Notes on scope

• External helper packages. A few notebooks import the authors' personal helper packages jctk / jcsctk (general single-cell utilities). These are not bundled here; the analyses can be followed without them.
• Illustrative examples. Where a step was repeated across many cell types or samples by different authors (e.g. per-cell-type cell-state annotation), a single representative notebook is included as an example rather than the full set; this is noted in the relevant directory README.
• Result tables. Large result tables (e.g. Hotspot modules, Milo/GSEA outputs) that are published as Supplementary Tables of the paper are not duplicated here.
• Data privacy. Internal compute paths, usernames and identifiers have been replaced with placeholders. Donor/sample codes that also appear in the paper and public data releases are retained.