qPCR Ct analysis and ΔΔCt done right — a reproducible pipeline

qPCR is where most gene-expression papers quietly become non-reproducible. The math looks simple — subtract two Ct values, raise 2 to the negative power — but every step hides a decision that, if undocumented, makes the final fold-change impossible to defend. This piece walks through a pipeline that survives both MIQE and ISO 15189.

What the MIQE guidelines actually require

MIQE (Minimum Information for Publication of Quantitative Real-Time PCR Experiments) is not optional if you plan to publish, and it is rapidly becoming the de-facto standard for clinical qPCR too. The non-negotiables:

• Assay efficiency per target, reported with slope, r², and y-intercept from a standard curve spanning ≥ 5 orders of magnitude.
• Reference gene validation per experimental condition using geNorm, NormFinder or BestKeeper. A single GAPDH across all conditions is a red flag.
• Ct threshold method declared and consistent — fixed threshold, second-derivative maximum, or automatic, but not mixed.
• No-template controls on every plate, with an explicit Ct cut-off policy (typical: NTC ≥ 40 or undetermined; sample Ct ≥ NTC − 3 cycles = flagged).

The reproducibility failure modes

1. The "baseline drift" gotcha

Different instruments and different software versions compute baseline differently. A Ct of 28.4 on a QuantStudio 7 with automatic baseline can be 28.1 on the same well analyzed in Design & Analysis 2.6 with manual baseline cycles 3–15. If you do not lock baseline settings in your SOP, six months from now you will not be able to reproduce your own paper.

2. Reference gene instability across conditions

A stress condition that changes GAPDH expression by 1.5-fold flips every ΔΔCt in the experiment. Run geNorm across candidate genes for each experimental arm — stable in control ≠ stable under treatment. Report the M-value and pairwise variation V2/3.

3. Efficiency correction is not optional

Pfaffl correction (2 → E) matters when target and reference have different efficiencies. A 90% vs 100% efficiency gap over 10 cycles introduces a 2.6-fold systematic bias in your ΔΔCt. If you are not using Pfaffl you are implicitly assuming 100% / 100% — document it.

The pipeline we use

1. Ingest — export the platform CSV with Ct, Tm, amp-status per well. Accept raw fluorescence when available (better baseline control).
2. QC gates — flag: no-amp wells, double peaks in melt curve, Ct SD > 0.5 among technical replicates, NTC within 3 cycles of sample, standard curve r² < 0.99, efficiency outside 90–110%.
3. Reference gene selection — geNorm + NormFinder on all candidates, output M-value ranking, auto-pick best pair or trio.
4. Normalization — ΔCt per sample using geometric mean of chosen references (not arithmetic — geometric is mandatory for log-scale data).
5. Relative quantification — ΔΔCt or Pfaffl with explicit efficiency per target. Output fold-change with 95% CI from replicate variance.
6. Statistical test — on ΔCt values (log-normal), not on fold-change (log-normal ratios). Report p-value and effect size separately.
7. Absolute quantification — when doing viral load, use a certified reference material chain (WHO International Standard where it exists, e.g. HIV, HBV, SARS-CoV-2) and report copies/mL with measurement uncertainty.

Viral load — the extra discipline

Clinical viral load has specific gotchas:

• LoD verification per lot — manufacturer LoD is a starting point, not the final number. Run a 5-level probit with ≥ 20 replicates per level on your workflow.
• Log-linear reporting — always log10, never linear, and with total analytical error ≤ 0.5 log at the medically relevant decision point.
• Commutability — your internal controls must behave like patient samples across the dynamic range. A plasmid-based control that drops out at high Ct is not commutable.

How AiLabrix handles qPCR

Drop the platform export (QuantStudio, LightCycler, CFX). The pipeline runs MIQE QC gates, picks references via geNorm+NormFinder, applies Pfaffl with per-target efficiency, computes ΔΔCt with CI, and produces a PDF that includes amplification curves, melt curves, standard curve fits, reference-gene stability, and the full audit log. One file in, one signed report out. [email protected] for a run on your data.