Ct Values Above 35: Reliable or Noise?
A Ct value above 35 is not automatically noise, but it is in the danger zone where distinguishing real signal from background becomes genuinely difficult. The short answer: a Ct of 35–37 can be reliable if you see tight replicates (SD < 0.5 Ct), a clean single-peak melt curve (SYBR assays), and your NTC wells are negative or at least 5 Ct values higher. A Ct of 38+ with scattered replicates and a wobbly melt curve? That's noise wearing a lab coat.
The reason this range is so fraught isn't mysterious — it's math. At Ct 35, you're looking at roughly 30–100 starting copies in a typical reaction. At Ct 38–40, you're often in single-digit copy territory, where stochastic sampling dominates. Whether a given molecule happens to be in your 20 µL reaction or stuck to the side of the tube matters more than your primer design. Understanding where that threshold falls for your assay, on your instrument, is what separates defensible data from wishful thinking.
Why Late Ct Values Are Inherently Noisy
The fundamental issue is Poisson statistics. When your template is abundant — say, 10,000 copies — the sampling variation is negligible. But at 10 copies, the actual number captured in a reaction follows a Poisson distribution with a mean of 10. Some wells get 6 copies, some get 14. That's a ~1.2 Ct spread just from sampling alone, before you factor in pipetting error or reverse transcription variability. At 3 copies, some wells will get zero — this is why you see 2 out of 3 replicates amplifying and one flat line at these low levels.
On top of stochastic sampling, you get contributions from:
- Primer dimer formation. Even well-designed primers can produce dimers at high cycle numbers when there's essentially no competing template. SYBR Green doesn't care what it binds — a 60 bp dimer fluoresces just like your 120 bp amplicon. This is the single most common source of false-positive signal at Ct > 35.
- Genomic DNA contamination. If you're doing RT-qPCR and skipped the DNase step (or it didn't work well), even femtogram quantities of gDNA can produce a Ct in the high 30s for many primer sets.
- Cross-contamination. Aerosol carryover from high-concentration samples or plasmid stocks. If your positive control is a plasmid at 10⁹ copies/µL and you're pipetting NTCs on the same bench, a Ct of 37 in the NTC is not surprising.
The instrument itself also matters. A QuantStudio 5 and a CFX96 handle baseline subtraction and threshold setting differently. Automatic thresholding algorithms can place the threshold in a region where late-cycle baseline drift gets called as amplification. This is particularly common on older LightCycler 480 instruments with aged filter sets or LED arrays — background fluorescence creeps up, and cycles 36–40 start looking like shallow amplification curves when they're really just drift.
How to Tell if Your Late Ct Is Real
There's no single test, but a checklist approach works well:
Replicate concordance. Three technical replicates should have an SD < 0.5 Ct. If your triplicates come back as 35.2, 36.8, and undetermined, that's a stochastic pattern consistent with <10 copies — reportable only with heavy caveats, if at all. If they're 35.4, 35.6, and 35.8, you have something.
NTC separation. Your no-template controls must be negative (truly undetermined at cycle 40) or at minimum 5 Ct values above your sample. A sample Ct of 36 with an NTC Ct of 39 gives you only 3 Ct of separation — that's an 8-fold difference, which sounds like a lot but is uncomfortably thin when you consider run-to-run variability.
Melt curve analysis (SYBR assays). A single sharp peak at the expected Tm for your amplicon is encouraging. A broad peak, a shoulder, or a peak at a lower Tm (often 76–80°C for primer dimers vs. 82–88°C for a typical amplicon) tells you the signal is at least partly artifactual. Run your product on a 2% agarose gel if you're unsure — 20 minutes and you'll have a definitive answer.
NRT controls. No-reverse-transcriptase controls are the overlooked hero of RT-qPCR. If your NRT gives a Ct of 37 and your sample gives a Ct of 36, your signal is predominantly genomic DNA, not transcript. You need at least 3.3 Ct of separation (10-fold difference) between your +RT and NRT to confidently attribute signal to RNA.
Amplification curve shape. Real amplification produces a sigmoidal curve with a clear exponential phase and plateau. At Ct > 35, you often get curves that barely rise above baseline and never plateau — more of a gentle upward drift than a true amplification curve. Most analysis software will still assign a Ct to these, but you shouldn't trust it.
Setting a Rational Cutoff for Your Assay
The popular advice of "throw out everything above Ct 35" is too blunt. The right cutoff depends on your assay's validated performance.
Run a standard curve down to single-digit copies. If you have a plasmid or synthetic standard for your target, prepare a dilution series that goes from 10⁶ down to 10, 5, and 1 copy per reaction. Run at least 6 replicates at each of the low concentrations. You'll see the point where replicate concordance falls apart and detection probability drops below 100%. That's your assay's limit of detection (LOD). For most well-optimized assays with 95–105% efficiency, this falls around Ct 34–37, but it varies.
Determine your limit of quantification (LOQ). This is the lowest concentration where your CV on copy number estimates is still acceptable — typically <25% CV. The LOQ is almost always a higher copy number (lower Ct) than the LOD. You might detect 5 copies at Ct 36, but you can only quantify reliably down to 50 copies at Ct 33.
This distinction matters for how you report results:
- Below LOQ but above LOD: you can say "detected" but shouldn't assign a meaningful fold-change.
- Below LOD: report as undetected, not as the Ct value the software spit out.
For relative quantification (ΔΔCt), the math amplifies errors at high Ct values. If your reference gene (ACTB) has a Ct of 18 with SD 0.15, and your GOI has a Ct of 36 with SD 0.8, your ΔCt is 18 ± 0.81 (propagated error). That's nearly a 2-fold uncertainty on the ΔCt alone, which becomes a ~1.8-fold uncertainty on your fold-change — before you even compare to a control group. The Livak (2⁻ᐩᐩCt) method (Livak and Schmittgen, 2001) assumes equal efficiencies and reasonably precise Ct values; it was never designed to handle this level of noise.
Practical Recommendations
If your GOI consistently shows Ct > 35 across all samples, consider whether the target is genuinely expressed at a biologically meaningful level in your system. Very low-abundance transcripts (some cytokines, signaling receptors in off-target tissues, or splice variants) legitimately live in this range, but you may need to switch approaches: increase cDNA input (if your reference gene can tolerate it without shifting below Ct 10), use a preamplification step (TaqMan PreAmp or similar), or move to digital PCR for true absolute quantification at low copy numbers.
If only some samples are above 35, those may represent real biological variation — a near-absent transcript in a control group that's induced in a treatment group, for instance. In this case, report the high-Ct samples as "detected but below LOQ" and consider a binary detected/not-detected analysis alongside your fold-change data.
For TaqMan assays, the primer-dimer problem is largely eliminated because the probe provides sequence-specific detection. This makes Ct values of 35–37 somewhat more trustworthy than SYBR equivalents, though stochastic sampling noise still applies. If you're routinely working in this range, TaqMan (or other hydrolysis probe chemistries) is worth the extra cost per target.
For NTC wells that show late amplification (Ct 37–40), don't panic, but don't ignore it either. In SYBR assays with well-designed primers (checked via melt curve), NTC amplification at Ct > 38 is usually primer dimer and not a contamination crisis. But if your NTC Ct is creeping down over successive runs — 39, then 37, then 35 — you have a progressive contamination issue, likely from amplicon carryover, and you need to clean your workspace, remake reagents, and aliquot fresh primer stocks.
A good workflow for any experiment with targets near the detection boundary: run your data through systematic checks for replicate concordance, NTC separation, and melt curve quality before interpreting fold-changes. VoilaPCR flags samples with Ct > 35 automatically and checks them against your NTC values, so you can quickly see which late-Ct results are defensible and which should be excluded — without building a spreadsheet from scratch every time.
The bottom line: Ct 35 isn't a magic number. Your assay's actual LOD is the number that matters, and the only way to know it is to measure it.