snapatac2.tl.diff_test#

snapatac2.tl.diff_test(data, cell_group1, cell_group2, features=None, covariates=None, direction='both', min_log_fc=0.25, min_pct=0.05, solver='lbfgs')[source]#

Test regions for differential accessibility between two cell groups.

Use this function to compare two explicit cell sets with logistic-regression likelihood-ratio tests after filtering by detection fraction and fold change.

Anti-Patterns#

  • Do NOT pass group names directly; pass cell indices, cell barcodes, or a Boolean mask for each group.

  • Do NOT use covariates; the parameter is currently not implemented.

  • Do NOT expect features failing min_pct or min_log_fc to appear in the output table.

param data:

Annotated data object with cells in observations and regions in variables.

type data:

AnnData | AnnDataSet

param cell_group1:

Cells in group 1 as indices, barcodes, or a Boolean mask.

type cell_group1:

list[int] | list[str]

param cell_group2:

Cells in group 2 as indices, barcodes, or a Boolean mask.

type cell_group2:

list[int] | list[str]

param features:

Region names or indices to test. If None, test all regions that pass filtering.

type features:

list[str] | list[int] | None

param covariates:

Reserved for covariates; currently raises NameError when provided.

type covariates:

list[str] | None

param direction:

Direction of enrichment to retain. "positive" keeps regions enriched in group 1, "negative" keeps regions enriched in group 2, and "both" keeps either direction.

type direction:

Literal[‘positive’, ‘negative’, ‘both’]

param min_log_fc:

Minimum absolute or directional log2 fold change required before testing.

type min_log_fc:

float

param min_pct:

Minimum fraction of cells with nonzero accessibility in either group.

type min_pct:

float

param solver:

Solver passed to sklearn.linear_model.LogisticRegression.

type solver:

str

returns:

Differential accessibility table sorted by adjusted p-value, with columns "feature name", "log2(fold_change)", "p-value", and "adjusted p-value".

rtype:

polars.DataFrame

Examples

>>> import snapatac2 as snap
>>> adata = snap.datasets.pbmc5k(type="annotated_h5ad")
>>> group1 = list(range(50))
>>> group2 = list(range(50, 100))
>>> result = snap.tl.diff_test(adata, group1, group2, features=list(range(20)))
>>> set(result.columns) <= {"feature name", "log2(fold_change)", "p-value", "adjusted p-value"}
True
On this page