raisinfit

Python port of the RAISIN hierarchical generalized linear model for differential expression. raisinfit estimates mean expression and both cell-level and sample-level variance components given a sample × cell-type design, optionally correcting for batch with ComBat before fitting. The returned fit object is consumed by run_pairwise_tests, which runs the actual pairwise contrasts and emits volcano plots. Supports unpaired, paired, continuous, and custom designs.

Source: sample_clustering/RAISIN.py:202

Signature

def raisinfit(
    adata,
    sample_col,
    testtype="unpaired",
    group_col=None,
    individual_col=None,
    batch_col=None,
    sample_to_clade=None,
    custom_design=None,
    intercept=True,
    filtergene=False,
    filtergenequantile=0.5,
    n_jobs=None,
    verbose=True,
    seed=42,
)

Parameters

Name	Type	Default	Description
adata	AnnData	—	Single-cell or pseudobulk AnnData.
sample_col	str	—	Sample identifier column.
testtype	str	"unpaired"	One of "unpaired", "paired", "continuous", "custom".
group_col	str, optional	None	Column with grouping/feature variable. Takes precedence over sample_to_clade when present.
individual_col	str, optional	None	Subject column for paired designs.
batch_col	str, optional	None	Triggers ComBat before fitting.
sample_to_clade	dict, optional	None	{sample_id: group_label} — used when group_col is absent.
custom_design	dict, optional	None	Required when testtype="custom"; keys "X", "Z", "group".
intercept	bool	True	Include intercept in the fixed-effect design.
filtergene	bool	False	Drop lowly expressed genes before fitting.
filtergenequantile	float	0.5	Quantile threshold used when filtergene=True.
n_jobs	int, optional	None	CPU cores for parallel fitting (default: all cores).
verbose	bool	True	Print progress.
seed	int	42	Random seed for reproducible fitting.

Returns

dict — the fit object with keys:

Key	Meaning
"mean"	gene × sample expression mean matrix
"sigma2"	gene × group between-sample variance
"omega2"	gene × sample within-sample variance
"X"	fixed-effect design matrix
"Z"	random-effect design matrix
"group"	group assignment per sample
"failgroup"	groups where variance estimation failed
"sample_names"	ordered sample identifiers
"batch_corrected"	whether ComBat was applied

Usage

from sampledisco.sample_clustering.RAISIN import raisinfit
from sampledisco.sample_clustering.RAISIN_TEST import run_pairwise_tests

fit = raisinfit(
    adata=adata_cell,
    sample_col="sample",
    sample_to_clade=expr_clusters,
    testtype="unpaired",
    batch_col=None,
    intercept=True,
    n_jobs=8,
)

run_pairwise_tests(
    fit=fit,
    output_dir="sampledisco_demo_output/rna/raisin_results_expression",
    fdr_threshold=0.05,
)