iscc.data.scDNA ¶

scDNA(n_cells=100, **dna_kwargs)

Bases: DNA

Single-cell DNA-seq: per-cell low budget, small-kappa (lumpy amplification) depth nested in the run batch, Beta-Binomial alleles + explicit ADO + doublets.

Each cell gets its own DM draw (the per-cell amplification profile, the dominant single-cell noise) on weights ∝ that cell's copy number. ADO (one allele lost at a het locus) is a separate Bernoulli layer — the dominant single-cell allele artifact — applied before the Beta-Binomial alt-count draw. The single-cell mode_defaults set a small-kappa amplification regime (kappa=5 lumpy MDA/MALBAC) plus ado_rate=0.20, beta_binom_conc=30, doublet_rate=0.02, and coverage is scaled ~0.3x the bulk depth.

Parameters:

Name	Type	Description	Default
`n_cells`	`int`	Number of cells to sample and assay, capped at the number available. Ignored when `cell_subset` is passed to `run`.	`100`
`**dna_kwargs`		Forwarded to the `DNA` base. See `DNA` for the shared `breadth` / `data_mode` / depth / hyper-parameters.	`{}`

Methods:

Name	Description
`run`	Assay single-cell DNA (copy number and SNVs) for the sampled cells.
`to_anndata`	cells × loci AnnData: X = coverage; alt/vaf/true_cn/true_af/ado in layers.

run ¶

run(cell_data, cell_subset=None, **kwargs)

Assay single-cell DNA (copy number and SNVs) for the sampled cells.

Emits, per cell and per observed locus, the read depth, alternate-allele counts, and VAF, applying doublets, allelic dropout (ADO), and Beta-Binomial allele noise. In "binary" data mode a genotype call is produced in observed_snvs instead of counts.

Parameters:

Name	Type	Description	Default
`cell_data`	`dict`	Per-cell ground-truth tables from the sampling stage; uses the SNV table `cell_snv` and the copy-number ground truth.	required
`cell_subset`	`array-like of cell IDs`	Restrict the assay to these cells. By default every sampled cell is used.	`None`

Returns:

Type	Description
`scDNA`	`self`, with per-cell tables in `coverage`, `alt_counts`, `vaf`, and `true_cn`, and the QC/ground-truth table in `obs`. Call `to_anndata` or `write` to export.

to_anndata ¶

to_anndata()

cells × loci AnnData: X = coverage; alt/vaf/true_cn/true_af/ado in layers.

Technology presets¶

Single-cell DNA platforms differ mainly in the whole-genome-amplification (WGA) chemistry and depth. Set breadth for the locus set / depth regime, and override the amplification knobs — kappa (lower = lumpier amplification), ado_rate (allelic dropout), and mu_depth (coverage) — via the constructor. These are starting points; estimate_dna fits them from real data.

Platform	`breadth`	`kappa`	`ado_rate`	Depth	Notes
MALBAC (default)	`"wgs"`	~5	~0.20	shallow (~9×)	Quasi-linear WGA, moderate bias — the built-in single-cell defaults.
MDA	`"wgs"`	~2–3	~0.30	shallow	Stronger lumpiness and higher dropout than MALBAC.
DLP / DLP+	`"wgs"`	~500 (near-uniform)	~0.05	very shallow (low `mu_depth`; DLP+ is ~0.01–0.1× the genome)	Direct single-cell library, no WGA bias: near-uniform coverage, low dropout.
Mission Bio Tapestri	`"panel"`	~5–10	~0.15	very deep	Targeted single-cell SNV amplicon panel.
10x CNV (Chromium)	`"wgs"`	~50	~0.05	very shallow	Droplet shallow WGS for copy number.