qPCR

Two qPCR workflows ship in Gradiance. They answer different questions and need different inputs, but they share a parser for Ct values and the same QC checks.

The two workflows

• Relative expression (ΔΔCt). Compare a target gene across conditions, normalized to a reference gene. Output is fold-change (2^−ΔΔCt) per condition.
• Absolute quantification. Copy number from a standard curve of known templates. Output is copies (or copies/µL, copies/cell) for each unknown.

Relative ΔΔCt

The math

For each well: ΔCt = Ct(target) − Ct(reference). Then: ΔΔCt = ΔCt(sample) − ΔCt(control). Fold-change is 2^−ΔΔCt.

That formula assumes 100% PCR efficiency for both target and reference. If your efficiencies are off (typically 90–110%), ΔΔCt understates or overstates fold-change. Gradiance flags efficiency drift if standards are present; otherwise the fit proceeds with the standard 2^−ΔΔCt formula and notes the assumption.

What you need on the plate

• Target — the gene of interest, in technical triplicate per biological replicate.
• Reference — a stable housekeeping gene (GAPDH, β-actin, 18S rRNA, or whatever your lab has validated as stable in your system). Same triplicate structure.
• Calibrator — one condition designated as “control” or “untreated.” Every other condition is reported as fold-change relative to this.

Tell the agent which gene is the reference: “use GAPDH as reference” or include a column labelled reference in your data.

Significance testing

For two conditions, Gradiance runs a two-sample t-test on the ΔCt values (not on fold-change — the log scale is closer to normal). For more conditions, it uses one-way ANOVA with Tukey HSD. P-values are reported alongside the fold-change.

QC specific to ΔΔCt

• Replicate Ct spread. Technical replicates of the same well should agree within 0.5 cycles. Spread > 1 cycle is flagged.
• Reference instability. The reference Ct should be similar across conditions. If GAPDH shifts by > 1 cycle between treatments, the assumption that it's stable is breaking — pick a different reference or use multiple references and geometric mean.
• No-template controls. Should have no amplification (Ct undetermined or > 40). Late amplification in NTCs flags contamination.

Absolute quantification

The math

Fit a linear regression of Ct versus log₁₀(copies) using known standards. Interpolate copy number for each unknown from its measured Ct.

PCR efficiency is computed from the slope: E = 10^(-1/slope) - 1. A slope of −3.32 corresponds to 100% efficiency. Values between −3.1 and −3.6 (90–110% efficiency) are acceptable.

What you need on the plate

• Standards — at least 5 dilutions of a known template, ideally spanning 5+ orders of magnitude. In duplicate or triplicate.
• Unknowns — your samples. Triplicate per sample.
• NTC — no-template control. Same expected behavior as ΔΔCt: no amplification.

QC specific to absolute quant

• Standard curve R² > 0.98. Anything lower means the standards weren't prepared with consistent dilution factors, or contamination is leaking in at the low end.
• Efficiency 90–110%. Outside this range, the slope-derived copy numbers are off by enough that you should fix the assay before reporting.
• Unknowns within standard range. Extrapolated unknowns are flagged. Dilute or concentrate to land within the curve.

Designing a qPCR run that fits well

• Three biological replicates per condition, three technical replicates per biological. The biological reps drive your statistics; the technical reps suppress pipetting noise.
• Validate your reference gene in your specific system. The “always-stable” references in textbooks are not always stable in practice.
• Keep all wells for one comparison on the same plate. Cross-plate comparisons require an inter-plate calibrator and a known adjustment factor.
• Always run NTCs. Late or unexpected amplification in NTCs invalidates everything else on the plate.

What the result looks like

Relative runs return fold-change with 95% CI per condition, plus p-values from the appropriate test. Absolute runs return copy numbers per unknown with CI from the standard curve. Both come with the standard exports: figure (PNG/SVG/PDF), parameter table, and a Python script that recomputes everything from raw Ct data. See Understanding your results for how to read the result panel.