Repository navigation
Discussion: other formats worth supporting in adata convert #31
Description
Part of #20.
Candidates beyond the first round. Please 👍 or comment with the ones you actually run into (and example files, if you can share them).
Pipeline outputs
- STARsolo full
Solo.out(Gene / GeneFull / Velocyto / SJ) → one AnnData with layers — partly covered by Convert: MTX and 10x MTX folders, import and export (streaming) #24 - alevin-fry USA mode (
quants_mat.mtx+ spliced/unspliced/ambiguous → layers) - kallisto | bustools (
kb countmtx / loom / h5ad outputs) - 10x Multiome / ATAC
filtered_peak_bc_matrix.h5— Convert: 10x HDF5 (Cell Ranger*_feature_bc_matrix.h5), import and export #23 with the Peaks feature type - Space Ranger / Visium and Xenium
cell_feature_matrix.h5+cells.parquet/spatial/— matrix via Convert: 10x HDF5 (Cell Ranger*_feature_bc_matrix.h5), import and export #23, coordinates →obsm["spatial"]
Container formats
- MuData
.h5mu— extract a modality (/mod/<name>is an embedded AnnData) or merge several into one. Current SCENIC+ writes this.
R ecosystem
- Seurat
.h5Seurat(SeuratDisk, plain HDF5 — readable with h5py), both directions - SingleCellExperiment via HDF5Array / zellkonverter
.h5 .rdsis out of scope (not HDF5, needs R)
Other
BPCells directories, UMI-tools count tables, .soft.gz (GEO), Excel.
Activity
adata convertto and from h5ad/zarr #20Merge pull request #72 from cellgeni/feat/convert-mudata
Claptar commentedon Oct 23, 2026
MuData
.h5muis in 0.16.0 (#72), both ways:adata convert x.h5mu -o x.h5adwrites each modality as its own AnnData (x.rna.h5ad,x.atac.h5ad, ...;--mod NAMEfor one). The joint results — global obs columns, obsm (a WNN embedding), obsp (the joint graph), uns — are taken onto each modality's cells throughobsmap; where a name clashes, the modality's own is kept, with a warning..h5mufiles and MuData Zarr stores are read, muon's older files included (checked on mudatasets'minipbcite.h5muagainst mudata's reading).adata convert x.h5ad -o x.h5musplits var by feature type as muon does (Gene Expression →rna, Peaks →atac, Antibody Capture →prot), or writes the whole AnnData under--mod; current mudata reads the result (checked on 10x's pbmc3k Multiome matrix). A MuData Zarr output is not written.Still open from the list above: STARsolo full
Solo.out, alevin-fry, kallisto | bustools, Visium / Xenium spatial,.h5Seurat, SingleCellExperiment. 👍 or comment on the ones you need.