Scientific Basis for Genetic ID’s Limits of Detection and Quantification for PCR Analysis
Genetic ID has established a limit of quantification for its PCR analytical methods of 0.1%, and
a limit of detection of 0.01%. This document explains why these limits have been established,
and presents three reasons why 0.1% is the lowest scientifically defensible threshold for
quantification of genetically modified material.
The first reason is statistical, relating to sampling of DNA preparations for PCR analysis. The
limitations of the PCR analytical systems used for quantification of GM material in food
samples are such that the amount of DNA that can be introduced into the reaction places a
boundary on the limit of quantification and the limit of detection of these methods.
PCR is capable of detecting even a single GM target molecule, if it is present in the DNA
sample introduced into the PCR reaction. However, if a sample of corn genomic DNA actually
contains 0.01% GM material, the number of GM targets in a PCR reaction will on average be
very small, in the range of one to four molecules. This is because the total number of corn
genomes in a PCR reaction will only be ten thousand to forty thousand. 1 Statistically speaking,
if one takes ten samples from a DNA preparation whose actual GMO composition is 0.01% GM
material, the number of GM targets in any given sample could be zero or one or two or three or
four, or etc. Amplifying these samples will lead to results having substantial differences in
signal intensity. These differences will not be related to the actual GMO content of the original
sample but will be due to the statistical variations related to sampling of that DNA preparation.
Thus, differences in signal intensity cannot be correlated with quantitative differences in GMO
content in samples that contain low GM levels such as 0.01%. Thus, we have established 0.01%
as our limit of detection, but because quantitative differences in signal intensity are likely to be
due to statistical fluctuations, quantitative analysis is not possible in this concentration range. In
fact, the absence of a PCR signal in a single sample cannot be used as evidence that the DNA
preparation did not contain GM material at the 0.01% level, since a certain number of DNA
samples will fail to contain GM targets, even though such targets are present at an average
frequency of 1 in 10,000 (0.01%) in the DNA preparation. Operating to a quantitative threshold
of 0.1% avoids the problems encountered when operating at the 0.01% level. In a sample
containing 0.1% GM material, the average number of targets in a sample of 50 ng will be 10,
and the probability that the actual number of targets in any specific sample will be within 50%
of the average will be about 90%. Thus, we set our limit of quantification at the 0.1% level.
1
The size of the maize genome is about 4.5X109 base pairs. The amount of sample DNA introduced into PCR
reactions is normally around 25 to 75 ng. Under certain conditions (when there are no inhibitors of the PCR process
present in the DNA prep and the concentration of targets in that preparation are very low) the amount can be
increased, but only to around 200 ng. Using Avogadro’s number (6.023X10 23), one can calculate from these values
that the actual number of maize genomes present in a PCR reaction will range from around 10,000 (50 ng sample
DNA) to around 40,000 (200 g sample DNA). If the concentration of GM genomes in the sample is 0.01%, or 1 in
10,000, the number of GM genomes (the number of GM target molecules) would be 1 for a 50 ng DNA sample to 4
for a 200 ng DNA sample. Because the size of the soy bean genome is smaller than that of maize (2.5X10 9 base
pairs) Sampling statistics are a somewhat better for soy beans than for maize.
The second reason for setting the limit of quantification at 0.1% has to do with the inherent
technical limitations of the PCR process. It is a simple fact of any analytical system that the
precision of quantification is low near the limit of detection of that system. For PCR, as operated
for GMO analysis, that limit of detection is around 0.01%. If the limit of detection is 0.01%, it is
reasonable to set the limit of quantification ten-fold higher at around 0.1% to avoid the problems
with precision that occur near the limit of detection. Empirically, we have found that it is not
possible to make meaningful quantitative distinctions in the region of 0.01% GMO and in
general, below 0.1%. 0.01% corresponds to the presence of one or a very few DNA target
molecules in the PCR reaction. With so few molecules present, small inefficiencies in
amplification will lead to large differences in signal intensity. For instance, consider two PCR
reactions each containing two target DNA molecules. If during the first round of PCR both
targets are amplified in reaction number one, while only one of the two targets is amplified in
reaction number two (due to inefficiencies inherent in the PCR process), there would be a two-
fold difference in target quantity at the beginning of cycle number two. At the end of the PCR
process, this two-fold difference at the end of cycle number one would result in a two-fold
difference in signal intensity for the analysis as a whole, even though the number of target
molecules was identical in the starting reactions.
Both of the limitations described above restrict the ability to confidently and dependably
quantify GMO content at or near the 0.01% level. This is the case whether analysis is performed
using conventional PCR or Real Time PCR methods. It would be possible to compensate for
both of these limitations by carrying out many replicate analyses, and/or by increasing the PCR
reaction size and the amount of DNA introduced into the PCR reactions. However, adjustments
along these lines that would be of sufficient magnitude to allow reliable quantification at the
0.01% level would be impractical as part of a quality assurance system. These limitations also
lower somewhat the confidence with which it is possible to identify the presence or absence of a
specific GM crop variety or event below the 0.1% level. Therefore, Varietal ID analysis below
this threshold is more challenging.
The third line of reasoning that speaks in favor of a threshold of 0.1% is that the 0.1% standard
is technically the most stringent that can be applied to whole grains and beans, such as soybeans
and maize kernels. This limit is not due to the PCR DNA analytical method, but to practical
limitations in sample size for grains. This is ultimately linked to the number of individual units
(beans or kernels) that comprise a sample of these materials. The sample must contain a
sufficiently large number of kernels/beans to assure that the sample comprises a statistically
valid representation of the product lot as a whole. For instance, to reliably quantify at the 0.1%
level (1 soybean in 1000), sound statistics require that at least 10,000 beans, weighing about 2.5
kg, be ground into powder and thoroughly mixed, and duplicate samples of this homogeneous
powder subjected to DNA analysis. This is feasible, and we have shown that it is a practical
reality to operate to this threshold.
However, to move the threshold down to 0.01% would require a sample size of 100,000 beans,
weighing about 25 kg. Thus, even if the PCR analytical system would support a 0.01% threshold
(which points #1 and #2 argue against), this threshold would be impractical because it would be
prohibitive to ship and process grain/bean samples of this size for analysis.
The particle sizes of processed materials are much finer than those of whole grains/beans.
Therefore, in principle, sampling limitations would not preclude a more stringent threshold for
processed products. A smaller sample of flour, for instance, could be used and yet achieve
statistical representation of the flour lot, because even one gram of flour will contain millions of
particles. However, it would be impractical and self-defeating to set a more stringent standard
for processed material than for the unprocessed grain from which it is derived. One would
immediately encounter situations where grain, which was compliant with the highest threshold
practical with unprocessed grain (0.1%), might be rejected when converted into flour and
retested to the standard set for flour (for instance 0.01%). For this reason our Cert ID non-GMO
certification program is based on the thresholds achievable with whole grains.
We conclude the following:
1. PCR is capable of detecting a single or a very few GM target molecules, if present in the
reaction tube. Generally PCR reactions for GMO analysis contain around 10,000 DNA target
molecules. Therefore, under conditions commonly used for GMO analysis, the limit of
detection of the PCR method is around 1 in 10,000, or 0.01%.
2. However, three factors make it impossible to reliably use this 0.01% limit as a quantitative
threshold for GMO content. First, due to the very low number of GM targets molecules
present in DNA preparations that contain 0.01% GM material, replicate samples of such
DNA preparations will vary in the actual number of GM targets molecules present in a given
sample. This will lead to an unacceptable level of variability in the measured GMO content
of the replicates. Second, technical limitations in precision commonly found when operating
near the limit of detection of a method will also contribute to variability at the 0.01% level.
Third, size limitations for samples of whole grains make a threshold of 0.01% impractical.
Because of these three factors, we conclude that it is not feasible to operate to a threshold of
0.01%.
3. Operating to a threshold of 0.1% avoids all three of these problems. (1) The concentration of
GM targets present in samples containing 0.1% GM material is sufficient to avoid problems
of statistical variation during sampling. (2) Because 0.1% is ten-fold above the limit of
detection of the PCR method, technical limitations inherent in operating near the limit of
detection are avoided in this concentration range. (3) The sample sizes for whole grains that
are required to operate to a threshold of 0.1% are practically feasible. Thus, it is possible to
operate reliably and practically to a threshold of 0.1%.