omicverse.metabol.drift_correct

Contents

omicverse.metabol.drift_correct#

omicverse.metabol.drift_correct(adata, *, injection_order, qc_mask, frac=0.5)[source]#

Correct systematic signal drift using LOESS regression on QC samples.

Fits log1p(intensity) ~ injection_order per feature on QC samples only, then divides the raw intensities in both QC and real samples by the fitted drift curve.

Parameters:

  • injection_order (str | ndarray) – Name of a numeric column in adata.obs (or a 1-D array) that gives the run order of each sample.

  • qc_mask (str | ndarray) – Bool column/array marking QC pool samples.

  • frac (float (default: 0.5)) – LOESS smoothing fraction (statsmodels.lowess frac argument). 0.5 is robust for runs of <500 samples; tighten to 0.3 for larger runs.

Returns:

Intensity matrix corrected in place on a copy; uns['qc_drift'] records the per-feature LOESS fit for reproducibility.

Return type:

AnnData

Contents