Parse Biosciences split-pipe

split-pipe writes, for the combined run and for each sample it was given, DGE_filtered/ (cells) and DGE_unfiltered/ (every barcode), each holding count_matrix.mtx (cells x genes), all_genes.csv and cell_metadata.csv:

run/all-sample/DGE_filtered/   (all-well/ in older versions)
run/all-sample/DGE_unfiltered/
run/<sample>/DGE_filtered/ ...

Give the run folder, a sample folder or a DGE_* folder:

adata convert run -o parse.h5ad
adata convert run -o s1.h5ad --sample s1
adata convert run -o parse.h5ad --all-samples --unfiltered
adata convert run -o mouse.h5ad --genome mm10
adata convert run/all-sample/DGE_filtered -o parse.zarr

Flag	Meaning
--sample	The sample folder to read; the combined all-sample by default
--all-samples	Write one output per sample, OUT.<sample>.h5ad
--unfiltered	Read DGE_unfiltered/ rather than DGE_filtered/
--genome	Keep that genome’s genes, and the cells split-pipe called as that species

Reading gives what Parse’s scanpy recipe gives (read_mtx, then both tables with pandas, indexed by bc_wells and gene_name), checked against it in CI, except that integer counts stay integers and features are named by gene id unless --var-names gene_symbols is given. obs holds every cell_metadata.csv column, typed as pandas reads it and stored as anndata writes it: a text column that repeats a value is a categorical. var holds gene_symbols (or gene_ids) and genome. --var-names and --make-unique apply as for 10x input.

--genome works on runs made with a reference of several genomes built by split-pipe, where each gene and cell names its own. A combined reference built for Cell Ranger (GRCh38-mm10-2020A) is one genome to split-pipe, so every gene and cell names it, and the species shows only in the gene ids’ prefixes.

Both files are read in two streaming passes, as for 10x MTX folders, so a run’s size is not a limit: the first checks the matrix against both tables before any output exists, and the second writes obs and X.