cnverse

Harmonise the outputs of heterogeneous CNV/CNA callers into one unified AnnData, so downstream exploration and visualisation can be written once.

cnverse is a thin interoperability layer, not another CNV caller. You run XClone / Numbat / CalicoST / inferCNV / CopyKAT yourself; cnverse reads each one's output and returns an AnnData with a consistent layout.

import cnverse as cnv

adata = cnv.io.read_xclone("path/to/xclone_out/")
cnv.schema.validate(adata)
# -> AnnData (cells x segments) with a canonical schema, ready to explore

Supported readers

The allele-aware readers expose their tool-specific column names as constants at the top of each module — edit those to match your installed version rather than rewriting the reader.

Canonical schema (v0.1)

• X — primary continuous CNV signal; its meaning is recorded in uns["cnverse"]["x_meaning"].
• var — genomic features (gene | segment | bin) with chrom, start, end, feature_type.
• obs — clone, clone_prob, ploidy_status, cnverse_tool.
• layers — any of total_cn, major_cn, minor_cn, cnv_state, baf, rdr, prob.
• obsm["spatial"] — coordinates for spatial modalities.
• uns["cnverse"] — provenance + cnv_states legend; uns["clone_tree"] for tools that infer a phylogeny.

See cnverse/schema.py for the authoritative definition.

Install (dev)

pip install -e ".[dev]"
pytest

Adding a tool

Create src/cnverse/io/<tool>.py with a read_<tool>(...) that returns cnverse.schema.make_cnv_anndata(...), then re-export it from src/cnverse/io/__init__.py.