Readers and estimators¶

Readers¶

from_anndata ¶

from_anndata(
    adata,
    *,
    layer: str | None = None,
    missing_value: float | str = "nan",
    obs_levels: list[str] | None = None,
    var_index: str | None = None,
    transpose: bool = True,
) -> pd.DataFrame

Convert an AnnData object to the features x samples DataFrame mismap-qc expects.

AnnData stores data as obs (rows, typically samples or cells) x var (columns, typically features). mismap-qc wants the inverse. With transpose=True (the default), the output is features x samples.

Parameters:

Name	Type	Description	Default
`adata`	`AnnData`	Input. Requires anndata to be installed (`pip install mismap-qc[anndata]`).	required
`layer`	`str`	Layer name to use. None => adata.X.	`None`
`missing_value`	`float or str`	How to treat missing values: - `"nan"` (default): keep NaN as missing. - `"zero"`: treat exact zeros as missing. - float: treat values below this threshold as missing (useful for log-intensity matrices with a noise floor).	`'nan'`
`obs_levels`	`list of str`	obs columns to use as additional MultiIndex levels on the sample axis. Example: `obs_levels=["batch", "condition"]` produces a 3-level MultiIndex with levels [batch, condition, sample].	`None`
`var_index`	`str`	var column to use as feature names. None => adata.var_names.	`None`
`transpose`	`bool`	If True (default), output is features x samples. Set False if your AnnData is already in features x samples orientation (rare).	`True`

Returns:

Type	Description
`DataFrame`	features (rows) x samples (columns). When `obs_levels` is set, columns are a MultiIndex with the obs levels and the sample name as the innermost level.

Raises:

Type	Description
`ImportError`	If anndata is not installed.
`ValueError`	If `layer`, `obs_levels`, `var_index`, or `missing_value` is invalid.

Examples:

>>> df = from_anndata(adata)
>>> df = from_anndata(adata, obs_levels=["Batch", "Condition"])
>>> df = from_anndata(adata, missing_value=0.0, var_index="gene_symbol")

Estimators¶

estimate_lod ¶

estimate_lod(
    df: DataFrame,
    *,
    method: str = "min",
    quantile: float = 0.05,
    min_present: int = 3,
) -> pd.Series

Estimate a per-feature limit of detection from observed values.

For each feature, returns either the minimum observed value (method="min") or a low-quantile observed value (method="quantile"). Features with fewer than min_present observations return NaN (cannot estimate).

Parameters:

Name	Type	Description	Default
`df`	`DataFrame`	features (rows) x samples (columns). NaN = missing.	required
`method`	`str`	`"min"` (default) or `"quantile"`.	`'min'`
`quantile`	`float`	Quantile in [0, 1] used when `method="quantile"`. Default 0.05.	`0.05`
`min_present`	`int`	Minimum non-NaN observations required to estimate. Features below this threshold return NaN.	`3`

Returns:

Type	Description
`Series`	Per-feature LOD estimate, indexed by feature name.

Raises:

Type	Description
`ValueError`	If `method` is not one of {"min", "quantile"} or `quantile` is outside [0, 1].

Examples:

>>> lod = estimate_lod(df)
>>> lod = estimate_lod(df, method="quantile", quantile=0.1)