thebench.bio #1 Ecosystem for the Bio-Tech Industry
All PCR & qPCR questions

Question 18 of 20 Advanced Practical

How would you determine the limit of detection and limit of quantification of a qPCR assay?

Short answer

Run many replicates of a dilution series around the expected limit. The limit of detection is the lowest level giving at least 95 percent positive replicates, and the limit of quantification is the lowest level measured with acceptable precision, for example a coefficient of variation of 35 percent or less.

Why

Classical analytical methods estimate detection limits from the spread of blank signals. That does not work for qPCR, because a blank gives no Cq at all rather than a small number. Forootan and colleagues therefore define the limit of detection as the concentration that gives at least 95 percent positive replicates.

Under ideal conditions, with only Poisson sampling noise, that is about 3 molecules per reaction. MIQE 2.0 repeats that the limit cannot fall below 3 copies. Real samples do worse, because extraction, reverse transcription and inhibition add noise.

The limit of quantification is about precision. It is the lowest level whose back-calculated concentration is reproducible enough for the decision at hand. A 35 percent coefficient of variation is one criterion used in the literature, not a universal rule.

Both limits apply to the whole method in a given matrix. A limit set on clean plasmid in water overstates how well the assay performs in plasma or soil.

How to work it through

  1. Prepare a standard that matches the target, ideally quantified by digital PCR, and spike it into the real sample matrix.
  2. Make a fine dilution series spanning from clearly detectable down to mostly negative, for example 50, 20, 10, 5, 3, 2 and 1 copies per reaction.
  3. Run many replicates at each level, at least 10 and preferably 20, across more than one day and operator.
  4. Count the fraction of positive replicates at each level and take the lowest level at or above 95 percent as the limit of detection, or fit a probit model.
  5. Back-calculate concentrations for positive replicates and compute the coefficient of variation at each level.
  6. Take the lowest level that meets the predefined precision criterion as the limit of quantification.

An example

Twenty replicates are run at each level. At 10 and 5 copies all 20 are positive. At 3 copies, 19 of 20 are positive, which is 95 percent.

At 2 copies 17 of 20, and at 1 copy 12 of 20. Those fractions sit close to Poisson expectations of 95, 86 and 63 percent, so sampling noise dominates. The limit of detection is 3 copies per reaction.

The coefficient of variation of back-calculated quantities is 18 percent at 20 copies, 31 percent at 10 and 52 percent at 5. With a 35 percent criterion, the limit of quantification is 10 copies per reaction.

Where the answer stops holding

The estimate can be no finer than the dilution steps chosen. Twenty replicates still leave wide uncertainty around a 95 percent proportion. Limits measured in one matrix, extraction kit or instrument do not transfer to another without verification.

For regulated use, the replicate numbers and criteria come from the applicable guideline rather than from a paper.

A strong answer covers

  • Replicate positivity at 95 percent defines the detection limit
  • Poisson floor of about 3 copies per reaction
  • Quantification limit defined by a precision criterion
  • Measure in the real matrix across days and operators
  • Report both limits with the dilution levels tested

Common mistakes

  • Using the blank-plus-three-standard-deviations formula on Cq values from blanks that never amplify
  • Declaring the limit of detection as the last dilution that gave any positive well
  • Validating limits on standards in water and applying them to extracted clinical samples