mismap-qc

Missing-data validation for proteomics and RNA-Seq experiments.

It detects outlier samples, classifies dropout mechanism (MNAR versus MAR), tests for batch effects, and gates pipelines on configurable QC rules. Every check has a matching plot for when you want to see the problem rather than just check it.

Install

pip install mismap-qc

Optional extras: [interactive] for HTML matrices (plotly), [anndata] for AnnData input, [upset] for missing_upset().

What it validates

Check	What it catches
Sample completeness	Samples with too few detected features
Outlier detection	Samples with anomalous missingness against their group peers
Missingness mechanism	Dropouts driven by low abundance (MNAR) against random (MAR)
Batch effects	Features whose detection differs between conditions
Run order drift	Instrument degradation over a long acquisition

The shape of the API

Three entry points, in increasing order of strictness:

from mismap_qc import qc, assert_qc

report = qc(df, group_level="condition")     # always returns a MismapReport
report.passes(thresholds={...})               # bool, raises nothing
assert_qc(df, thresholds={...})               # raises MismapQCFailure

qc() returns a frozen MismapReport holding pandas DataFrames you can drill into. assert_qc() is the pipeline gate.

See Quickstart for a worked example, or the Tutorial for real CPTAC proteomics data.

Why this instead of something else

• missingno does general missing-data visualization with no omics awareness: no groups, no MultiIndex sample annotations, no MNAR mechanism.
• protti (R) classifies missingness mechanism but has no Python equivalent.
• great-expectations validates tabular data but does not understand missingness mechanism or omics-specific patterns.

mismap-qc covers all three through one API and reads AnnData natively.

Input format

A pandas DataFrame with features as rows, samples as columns, and NaN meaning missing or not detected. When the columns are a MultiIndex, the level names become annotation strip labels automatically.