CFX Maestro to Publication Figure in 60 Seconds

The whole trip from a finished CFX96 run to a figure you can paste into a lab meeting slide is three steps: export the Cq table from CFX Maestro, drop it on an analyzer that understands Bio-Rad's column names, pick your reference gene and control group. Everything else, from replicate averaging to the Welch t-test on ΔCt values, is arithmetic a computer should be doing while you get coffee.

Below is the exact export that works, what to look for in the file before you upload it, and the two settings that decide whether your fold changes are right. If you have ever rebuilt the same ΔΔCt spreadsheet for the fourth time because a collaborator added one more gene, this is the post for you.

The one export that works: Export All Data Sheets

CFX Maestro offers several exports and most of them are traps. The pivot-style "Quantification Summary" is readable by humans and useless to software. "Gene Expression" results depend on settings you chose inside Maestro and bake in a normalization you may not want. The export you want is the raw per-well Cq table.

1. Open the run in CFX Maestro (.pcrd file).
2. Export → Export All Data Sheets → Excel 2007 (.xlsx), or CSV if you prefer. Maestro writes one file per sheet; the one you need is named …Quantification Cq Results.xlsx (or .csv).
3. Open it once to sanity-check. The header row should be Well, Fluor, Target, Content, Sample, Biological Set Name, Cq, Cq Mean, Cq Std. Dev, …. If you see Cq spelled out as a column and one row per well per fluorophore, you have the right file.

Two things to check before you move on. First, the Sample column must actually contain sample names. If the plate was run without a plate layout, Sample is blank and Content just says Unkn, and no software can recover what you forgot to type in. Fix the layout in Maestro (Plate Setup → edit, then re-apply) and re-export. Second, Target must carry gene names, not fluorophore names. A Target column that says SYBR for every row means the plate was set up by fluorophore only.

The .pcrd file itself is encrypted and cannot be read by anything but Maestro. If you would rather skip the Excel step entirely, File → Save As and switch the type to RDML; that open XML format carries samples, targets and Cq values together and uploads directly.

What "60 seconds" looks like in practice

Drop the Cq Results file on VoilàPCR. The Bio-Rad parser recognizes the Fluor/Content/Cq header shape without you telling it anything about the instrument, maps Content codes to roles (NTC wells become no-template controls automatically, Std wells become standards), and lists your targets and samples.

Then two choices decide everything:

Reference gene. Pick the housekeeping gene you ran on the same plate. If you ran two (say GAPDH and HPRT1), the geNorm option in the method selector will rank their stability and use the geometric mean, which is what MIQE asks for. With one reference gene, ΔΔCt it is.

Control group. Your sample names decide how much work this is. Names like Control_1, Control_2, Treated_1 are split on the trailing replicate number into groups called Control and Treated, and Control is offered as the baseline. Names like S1 through S12 are just twelve samples; you will group them by hand in the group builder. Thirty seconds of naming discipline when you set up the plate saves five minutes here, every time.

Click Analyze. The result is the fold-change table per target and group, a bar chart with error bars drawn from the biological replicates, and a QC panel that flags the things Maestro never tells you about: an NTC that amplified at Cq 36, a replicate pair that disagrees by more than 0.5 cycles, a target whose Cq values sit above 35 where the Poisson noise starts to dominate.

The three settings people get wrong

Technical versus biological replicates. If each biological sample was pipetted in triplicate, those three wells must be averaged into one value before any statistics. The default handles this when the three wells share a sample name. If you named the wells Mouse1_a, Mouse1_b, Mouse1_c, the suffix stripping will see three different samples and inflate your n from one to three. Name technical replicates identically; only biological replicates get distinct names.

Efficiency. ΔΔCt assumes both target and reference amplify at 100 percent efficiency. If you ran standard curves and know the real efficiencies (anything from 90 to 110 percent is normal; a 94 percent target against a 103 percent reference is a 20 percent error in fold change across a 5-cycle ΔCt), switch the method to Pfaffl and type them in. If you did not run standard curves, ΔΔCt is the honest choice.

Statistics on the right scale. The t-test runs on ΔCt values, which are approximately normal, not on fold changes, which are log-normal. Bar heights on the figure are the geometric means back-transformed to fold change, and the error bars are the standard error on the log scale. If a reviewer asks why the error bars are asymmetric, that is the answer, and it is the correct one.

Exporting the figure

The graph exports as PNG at 300 DPI or as SVG for Illustrator, with a journal preset that sets fonts, bar widths and the significance brackets the way Nature or Cell figure guidelines ask. The data table exports as CSV, and the whole thing exports as a GraphPad Prism file if your lab's figure pipeline runs through Prism. The methods paragraph is generated from the settings you actually used, so the sentence "Relative expression was calculated by the ΔΔCt method (Livak and Schmittgen, 2001) normalized to GAPDH with the untreated group as calibrator" reflects the analysis, not a template.

The first analysis on your own data runs without an account. If the CFX export in your downloads folder is older than this morning, drop it on VoilàPCR now and see what the figure looks like; it will take less time than reading this paragraph did.