Guides · BioStudioAI

Flow Cytometry Data Analysis: Gating, Compensation & Controls

Last reviewed Author: BioStudioAI

Answer first. Fluorophores spill into neighbouring detectors. Compensation subtracts that spillover using single-stained controls — beads or cells, one per fluorophore, plus an unstained control, all at least as bright as your samples. Analyse nothing until compensation is applied and checked: plot each fluorophore against its neighbours on single-stained controls and confirm populations centre where they should. Over- and under-compensation both invent rare populations that do not exist.

1. Compensation comes first

Fluorophores spill into neighbouring detectors. Compensation subtracts that spillover using single-stained controls — beads or cells, one per fluorophore, plus an unstained control, all at least as bright as your samples. Analyse nothing until compensation is applied and checked: plot each fluorophore against its neighbours on single-stained controls and confirm populations centre where they should. Over- and under-compensation both invent rare populations that do not exist.

2. The controls that make gates honest

  • FMO (fluorescence-minus-one) controls set the boundary between negative and positive for spread-prone channels. An isotype control is not a substitute.
  • Viability dye on every sample: dead cells bind antibodies non-specifically and light up everywhere.
  • Unstimulated / biological negative controls define what “off” looks like in your actual system.
  • Reference or batch control when samples span days, so instrument drift is visible instead of absorbed into your groups.

3. A gating hierarchy in the right order

  1. Time gate — remove pressure spikes and clogs (events vs time should be steady).
  2. Singlets — FSC-A vs FSC-H (or -W); doublets fake double-positive cells.
  3. Live cells — viability-dye negative.
  4. Lineage gate — e.g. CD45+, then your lineage markers, in progressively narrower steps.
  5. Population of interest — boundaries set from FMO controls, applied identically to every sample.

Lock the template before you look at group differences, and apply the same gates to all files in the experiment. Moving a gate per sample “because this one looks different” is how unconscious bias enters — the gate position is a measurement decision, so it gets made blind to condition wherever possible.

4. Report populations, not just percentages

A percentage is a ratio; it moves when the denominator moves. Report, where you can: percentage of a clearly named parent, absolute counts (with counting beads or a recorded volume), and median fluorescence intensity on a stated scale. State the parent of every percentage in text — “% of live CD4+ T cells”, never a bare “% positive”. For MFI comparisons across days, anchor to your batch control.

Minimum publishable record: instrument and configuration, compensation controls used, gating hierarchy with a representative figure, FMOs for key boundaries, event counts per gate, and the exact parent of every reported percentage. If a reviewer cannot rebuild your gates from the record, the analysis is not finished.

Frequently asked questions

What is an FMO control and when do I need one?

An FMO contains every fluorophore except one, showing the spread the other dyes create in that empty channel. Use FMOs wherever a population boundary is not obvious — especially dim or continuous markers.

How many events should I collect?

Enough that your rarest reported population has a stable count — a common floor is 100 events in the final gate. Percentages built on 12 events carry enormous sampling error.

Can I fix bad compensation during analysis?

You can recalculate compensation from the stored single-stained controls, and you should if the matrix is wrong. What you cannot fix later is missing controls: unrecorded FMOs or viability cannot be invented after acquisition.

From raw file to finished report

BioStudioAI reads the raw instrument file, runs QC, statistics and figures, and writes the report and slides — every number computed by its engines, never typed by an AI. Hosted on your servers.

Open BioStudioAI See the platform

Frequently asked questions

What is an FMO control and when do I need one?

An FMO contains every fluorophore except one, showing the spread the other dyes create in that empty channel. Use FMOs wherever a population boundary is not obvious — especially dim or continuous markers.

How many events should I collect?

Enough that your rarest reported population has a stable count — a common floor is 100 events in the final gate. Percentages built on 12 events carry enormous sampling error.

Can I fix bad compensation during analysis?

You can recalculate compensation from the stored single-stained controls, and you should if the matrix is wrong. What you cannot fix later is missing controls: unrecorded FMOs or viability cannot be invented after acquisition.

Do this analysis in BioStudioAI

Upload the raw file and BioStudioAI fits, quality-checks and reports it with an audit trail.

See the platform →

Related

Setting up compensation controls that actually work

Compensation controls must be at least as bright as samples, use the same fluorochromes and tissue background, and be run for every fluorochrome plus

Compensation and spectral unmixing explained

Compensation subtracts measured spillover between detectors; spectral unmixing fits full emission spectra. Both need single-stain controls brighter th

FMO controls: a practical guide with examples

An FMO contains every reagent except one, pricing the spread that channel inherits from all others. Gates for dim or continuous markers belong on FMOs

Flow cytometry