Data files FCS and image metadata

An FCS file is a list of events plus a dictionary. An image file is a grid of pixels plus a header. Either one fails the next reader when the dictionary is thin. The decision this selection guide supports is which fields to require in a specification and in the archive, so a second person can tell what the numbers meant. How those numbers arise in flow is in how a flow cytometer reads a cell. Why the pixels still need optical controls is in fluorescence microscopy without the myths.

Instruments that write the files are in the scientific instruments catalogue. The experiment may be part of the molecular biology pathway. File-format requirements belong in a quote request.

What to require of an FCS file

FCS, the Flow Cytometry Standard, is the interchange format the cytometry field actually uses. The International Society for Advancement of Cytometry is the society associated with that standard and with minimum reporting expectations. A specification should say "FCS export" and then name the contents, because a file can be technically FCS and still have empty stain names.

Ask for the cytometer name, the begin time, and a parameter record for every channel you paid for. Each parameter should carry a short name and a stain or fluorophore name a stranger can read. Scatter parameters should say whether they are height, area or width. Fluorescence parameters should name the detector and the dye. Bit depth and range should be present so a later program knows the scale. A spillover keyword, or an equivalent compensation record, should be present when a matrix was applied. If the software writes compensation only inside a private workspace, say that in the specification and archive the workspace too.

The event list is the data segment. It does not include your polygons if the software never wrote them there. Treat the gate hierarchy as a required companion file. Treat the single-stain FCS files as required companions of the matrix. A paper that shows a gated plot and archives none of these is not reproducible from the public record.

What to require of an image file

The pixels are not the measurement until the header says how big a pixel is and how the light was collected. Require micrometres per pixel in x and y. Require the z step if the file is a stack. Require channel identity: fluorophore, excitation line, emission band. Require exposure or pixel dwell, gain, offset if it matters, and binning. Require the objective's magnification and numerical aperture, and the immersion if that is how the pixel size was achieved.

OME-TIFF is one open-minded container class for those fields. Vendor files can be acceptable when a documented reader exists and the fields are actually filled in. A JPEG exported for a talk is a view, not an archive. The MicroscopyU fluorescence section and the Evident fluorescence primer explain the optical fields. Your specification says the software must store them.

A practical archive list

Store the raw file, the workspace or the processing script, the control files, and a text note that maps sample names to tube labels. Store the lot of the antibody if identity matters. Do not store only the final PNG. Disk space is cheaper than a lost matrix. If a power cut interrupted a run, store the partial file under a name that says it was interrupted, rather than renaming it as the completed sample.

Opening last year's file

Imagine a reviewer asking why a double-positive gate moved. The FCS file opens. The cytometer name and the time match the notebook. Parameter short names are present, and the stain names are blank, so one channel might be the cytokine or the viability dye. The spillover keyword is missing, and the workspace was never copied. Clouds are still visible. The boundary is not defensible. If the single-stain files exist and the voltages were written down, you can rebuild the matrix. If the voltages lived only on a whiteboard, you cannot. The image equivalent is a folder of pictures with no micrometres per pixel and no exposure. A nucleus is still visible. A scale bar and a brightness comparison are not. That is why a specification lists fields rather than the word data.

Minimum information for a flow experiment, the reporting idea discussed in the cytometry community, asks for the sample, the instrument configuration and the analysis in enough detail that another laboratory can understand the comparison. You do not need to memorise a checklist to apply that spirit. If a stranger cannot tell the trigger channel, the viability dye and the parent gate, the archive is short. If a stranger cannot tell the objective and the pixel size, the image archive is short. Write those into the enquiry as acceptance checks. A test file should be opened away from the instrument computer, and the fields should be visible there. Keep the antibody lot in a plain table if the format has no clean slot for it. An informal keyword is better than a missing lot, and the plain table should exist as well.

Failure modes when you specify poorly

A quotation that says "exports data" may mean a screenshot. Ask for the format by name. A cytometer that writes FCS without stain names forces every analysis to depend on a paper key you will lose. A microscope that writes a proprietary stack with no pixel size forces every scale bar to be reconstructed from memory. A cloud folder that keeps the final figures and drops the raw files fails the specification even if the instrument was perfect.

Opening the file once, on the instrument computer, does not prove a second program can read it. Before you accept a format, open a test file elsewhere. That single check is worth more than a long feature list.

Safety, privacy and research limits

Metadata can contain sample identifiers. If those identifiers are human, the archive is also a privacy question for the institution, separate from the optics. Biosafety of the material that was run is unchanged by a good file, and remains an institutional decision with the CDC BMBL as one reference. This guide does not approve unfixed human samples, and it does not make a data file a medical record. Do not put clinical claims in a research FCS keyword because the field was empty.

What to send with an enquiry

Ask for FCS with named parameters and a spillover record, an exportable gate workspace, and image files that store pixel size, channels, exposure, gain, binning and the objective. Say you will test-open a sample file. Use the scientific instruments catalogue and the quote request. Ask whether a quotation is possible. A software name without the field list is not a data specification.