Intelligence

Analysis

flowClean and Acquisition Artefacts: What a Changed Cell Composition Really Means

A 2016 flowClean study found fluorescence anomalies in 13.7% of 29,228 public FCS files. Its compositional changepoint method can flag suspect intervals, but does not by itself validate a clinical MRD result.

Technical Review of a peer-reviewed original study in Cytometry Part A (first published 18 March 2016). DOI 10.1002/cyto.a.22837. Original CFCM educational analysis; not an independently validated clinical assay.

Flow cytometry QC: what happens when fluorescence changes during acquisition?

A cytometer can acquire a coherent-looking population while a transient fluidics disturbance has shifted fluorescence. A single diagnostic dot plot may not expose every affected channel. flowClean was designed to identify acquisition intervals where the relative composition of many artificial cell populations changes, not merely to apply another static fluorescence gate.

The study's 13.7% is a screening observation, not a clinical failure rate

Fletez-Brant and colleagues examined 29,228 publicly available FlowRepository FCS files and identified fluorescence anomalies in 13.7%. This is a retrospective observation in a heterogeneous public collection. It is not the expected incidence of assay failure in an accredited clinical laboratory or proof that every flagged interval should be deleted. Their independent proof-of-concept and vaccine-study material included 684 flow files, typically drawn from three time points in 228 participants, and should not be merged into the public-file denominator.

How flowClean works: artificial phenotypes, then compositional changepoints

Each fluorescence parameter is partitioned above or below its median. Combinations of these partitions produce deliberately artificial phenotypes, not diagnostically identified cell types. In the reported configuration, informative groups required more than 500 events each, with at least five such groups, and acquisition was divided by default into 100 equal-event time bins. Relative population frequencies were transformed into centered log-ratio space, then checked for changepoints. A modified FCS file records the events from suspect time periods so the analyst can retain and review the original acquisition trail.

Why the threshold matters for rare populations

The more-than-500-events criterion was selected for stable estimation of group-frequency changes within their method. It does not mean an MRD population needs 500 abnormal events, nor does a cleaning decision prove that a rare residual population is absent. A rare population may simply be too sparse to drive this QC statistic. If removing a flagged interval changes the numerator, the denominator, or the fraction of viable and interpretable cells, reassess the result's sample-specific uncertainty rather than silently recalculating a percentage.

Original CFCM four-stage conceptual diagram explaining acquisition anomaly detection, event exclusion, review of file flags and clinical denominator; no study figure or patient events copied.
Original CFCM conceptual diagram. Educational synthesis; not clinical validation.

A practical review trail for clinical flow cytometry

Retain the original FCS, the event-time plot, the software version and parameters, the boundaries and event counts of every excluded interval, and side-by-side pre/post-exclusion frequencies for any population with clinical significance. Verify acquisition stability, fluidic disturbances, compensation and specimen adequacy independently. A compositional flag is a request for review; it cannot substitute for clinical LOD/LOQ, validated gating, or a defensible no-call policy.

What this original paper cannot establish

The source demonstrates a historical automated acquisition QC approach. It did not prospectively establish performance in routine AML, CLL or myeloma MRD, nor validate treatment decisions or the effect of quality-control filtering on very rare malignant populations. Report the paper's study type and the difference between flagging a period and declaring a specimen clinically acceptable.

Daily Learning Pill

Cleaning is a change to the measurement record. Ask whether the flagged intervals changed the effective event denominator and whether the observed rare-event conclusion is robust before and after exclusion.

Explore CFCM Troubleshooting and CFCM Resources for event-quality and gating principles. Read the original research article for complete methods and limitations.