Two important parameters that describe the quality of quantitative data are accuracy and precision.
- Accuracy is the degree that measurements equal the actual or true quantity in the theoretical absence of variability. An accuracy error is a consistent shift in the results away from the true quantity.
- Precision is the random variation of repeated measurements.
Improved precision in a qPCR experiment enables you to discriminate smaller differences in nucleic acid copy numbers or fold changes. If you run replicates, you can measure reproducibility and assess precision in your experiments.
Importance of precision
Precision is important for quantification, because variation impacts the meaningfulness of the results. If the variation is low, the results will be more consistent and statistical tests will have an improved ability to discriminate fold changes in gene quantities. If the variation is high, the results will be less consistent and statistical tests will have a reduced ability to discriminate fold changes in gene quantities. High variation may make it necessary to increase the number of replicates to improve discrimination power, which would increase cost.
Precision can also impact qualitative tests. Excessive variability may cause a true positive sample to be recorded as negative and vice-versa.
Coefficient of variation (CV) is a measure of precision and is equal to the standard deviation quantity divided by the mean quantity of a group of replicates. CV is often represented as a percentage. As a benchmark for system precision, Applied Biosystems brand real-time PCR instruments have an instrument performance specification maximum cut-off of 11% CV. Note that the performance verification test involves many technical replicates, e.g., 72 replicates in a 96-well plate.
Standard deviation (SD) describes a portion of a normally distributed population and is relative to the mean of the population. For example, a region of a normally distributed population bounded by plus and minus one standard deviation from the mean encompasses 68% of the population. SD is not a precision measure by itself, but it is very important for calculating CV.
Standard error (SE) is a measure of sampling error, providing upper and lower boundaries for how distant the measured mean is likely to be from the true mean of the population. SE is equal to the SD divided by the square root of the number of replicates. SE is not interchangeable with SD.
The two most common types of replicates used in real-time PCR are technical and biological.
- Technical replicates are repetitions of the same sample. The amplifications are performed in multiple wells using the same template preparation and the same PCR reagents. Technical replicates help protect the data, e.g., if one amplification fails, other wells may succeed. In addition, technical replicates offer a number of benefits, such as providing an estimate of system precision, improving experimental variation and allowing for potential outlier detection and removal. However, technical replicates add cost and reduce the throughput, so you must decide on an optimal number of technical replicates, balancing benefits vs. costs. In basic research, triplicates are a commonly selected replicate number.
- Biological replicates are different samples that belong to the same group. They are amplifications that use the same PCR reagents and similar (but not identical) samples for the template reagents. Biological replicates take into account variation within a defined group. For example, when examining the effect of drug treatment on the gene expression level of a mouse mRNA target, multiple mice are needed as samples to estimate the variation of that target in the population. In this example, a biological group might consist of a group of mice treated with the same amount of drug.
Replicates do cost more, occupy more space in the thermal cycler block, and consume more sample. You must weigh these factors against the need for precision.
Effect of replicate number on precision
Each sample aliquot assayed in real-time PCR provides an estimate of the gene target quantity in that sample. Due to the impact of random variation, the estimate will vary to some degree from the true value. Using the mean value from multiple aliquots will tend to reduce the impact of random variation. This effect is true for both technical and biological replicates.
For example, if an instrument performance verification test produced a precision value of 5% CV, the precision measured from smaller groups of replicates can be significantly higher or lower than 5%.
Types of variation
There are three variation types in a real-time PCR experiment or test.
- System variation is the variation inherent to the measuring system. Contributors to real-time PCR system precision include pipetting variation and instrument-derived variation. System precision can be estimated by assaying multiple aliquots of the same sample, called technical replicates.
- Biological variation is the true variation in target quantity among samples within the same group. The term “biological” is being used due to the prevalence of biological samples in real-time PCR experiments, but the concept of group-level variation can apply to other types of samples considered to be non-biologically derived. Biological or group variation is theoretically taken into account by assaying multiple samples belonging to the same group.
- Experimental variation is the variation measured for samples that belong to the same group. Experimental variation is used as an estimate of biological or group variation. Due to the influence of system variation, experimental variation will likely not be exactly equal to biological variation.
System variation can impact experimental variation by either increasing or decreasing it relative to the true biological variation. The larger the system variation, the larger its potential impact on experimental variation.
Absolute vs. relative quantification
Absolute quantification is measuring the target using a fundamental unit of measure. In the case of real-time PCR, the absolute unit is the RNA or DNA molecule. Absolute quantification allows single sample results to be meaningful. In addition, absolute quantification allows gene-to-gene and sample-to-sample quantitative comparisons.
Relative quantification is measuring the target using a non-fundamental or arbitrary unit of measure. Relative quantification does not allow single sample results to be meaningful, nor does it allow gene-to-gene quantitative comparisons, but sample-to-sample quantitative comparisons can be made.
Many real-time quantitative PCR experiments involve comparing target quantities between two or more biological groups, such as Control and Treated. If a fold change in target quantity was observed between two groups, an important question is whether that fold change could be reasonably accounted for by experimental variation (random chance). To assess that possibility, a statistical test is performed, such as the t-test.
Statistical tests produce one of two possible results: non-significant or significant.
- Non-significant result means that experimental variation could reasonably account for the observed fold change. The treatment might have caused a change, but the statistical test could not make that determination with sufficient confidence.
- Significant result means random chance could not reasonably account for the observed fold change and therefore in a well-controlled experiment, the test variable, the treatment, is the most likely explanation for the change.
Increasing the number of biological replicates and reducing variation allows the statistical test to discriminate smaller fold changes. Reducing variation might also allow the same discrimination power with fewer replicates, which would reduce the cost of the experiment.
Researchers studying biological samples should also consider physiological significance. With sufficient replicates and low variability, small fold changes might be assessed as statistically significant, but the change might not be large enough to significantly alter cellular metabolism. For example, in eukaryotic gene expression, a 2-fold change is often considered to be the minimum for physiological significance.
How to improve precision
You may or may not have much control over biological or group-based variation. For example, one approach to reducing biological variation is using in-bred strains of lab animals grown under controlled conditions, but this may not be an option in some circumstances.
You can take steps to improve system precision, which in turn will help make experimental variation a better estimate of biological variation.
- Ensure good instrument performance. All Applied Biosystems brand real-time PCR instruments have maintenance procedures to help ensure optimal performance over their lifespan. These procedures include temperature verification, calibrations and performance verification. Maintenance can be performed by a trained service engineer during planned maintenance visits, which are often part of service contracts. Inspect blocks periodically for cleanliness, as contaminating material can affect results, e.g. writing residue on reaction plates absorbing light.
- Test dynamic range. Dynamic range is the minimum and maximum sample amount that can be processed through an assay system, while still producing accurate and precise results. An assay system is the complete workflow from sample to results. Real-Time PCR instruments can support a large dynamic range. For example, a 20 mL PCR volume should provide a dynamic range potential of approximately 9 logs. However, a number of factors unrelated to the instrument, such as sample quality, target abundance and reverse transcription efficiency, will likely impose dynamic range limitations. Therefore, the dynamic range of a new assay system should be tested upfront using a small, representative set of samples. Without such testing, there is a risk samples may be processed outside of dynamic range, resulting in a loss of accuracy and precision.
- Technical replicates. Increasing the number of technical replicates will tend to reduce the effect of system variation. In addition, technical replicate variation should be monitored. Unusually large variation may indicate a problem.
- Biological replicates. Increasing the number of biological replicates will tend to reduce the effect of system variation. System variation can increase and decrease experimental variation. If the effect of system variation has caused experimental variation to be abnormally low, false positive statistical results could occur. Biological group variation should be monitored for abnormally low values and if found, that portion of the experiment should be repeated or discarded.
- Multiplexing. Multiplexing is the amplification and detection of multiple gene targets per well. If the assay used to normalize the data is present in the multiplex, normalizing the target data by the normalizer data from the same well will create a precision correction, improving precision.
- Passive reference dye. A passive reference dye is a dye present in the real-time PCR reaction at a fixed concentration that does not participate in amplification, but is used to normalize the reporter dye signals. A passive reference dye can improve precision in a number of ways, such as correcting for variations in assay master mix volume and correcting for optical anomalies.
- Good pipetting technique. Ensure each pipette is in good working order, the tips fit snugly and the pipettes are being operated according to manufacturer’s recommendations, including minimum volume. For multi-channel pipettes, check that volume levels are consistent across all tips. Pay special attention when pipetting liquids that are unusually viscous or have detergents.
- Twenty percent rule. If the sample volume exceeds 20% of the PCR reaction volume, an optical anomaly called “optical mixing” is likely to occur, which can harm precision. Optical mixing can be prevented by vortexing the sealed plate for a few seconds.
- Good plate loading technique. When loading the plate with reagents and samples, visually ensure consistent volume deliveries. After sealing the plate, centrifuge the plate to bring all liquids to the bottom of the wells and bring potential air bubbles that may be trapped underneath the liquid to the surface.
- Good analysis technique. You should gain familiarity with how to analyze real-time PCR data, e.g., designate the correct dyes, set baseline, set threshold, monitor for anomalies, etc. Thermo Fisher Scientific provides documents and videos that explain how to run and analyze real-time PCR experiments. Technical Support can provide remote, live support and Field Applications Specialists can provide on-site training. Training classes are also offered at Thermo Fisher Scientific facilities.
- Maintain block cleanliness. Writing on the 96-well plate, optical caps, or optical tubes is not advisable. A qPCR detector is a camera that collects fluorescent signals. Using ink to label the PCR plate surface can attribute additional fluorescent signal to what is already being collected. Instead, note the contents of each well on a sheet of paper, or spreadsheet, with a 96-well pattern to keep track of pipetting. Many compounds found in laboratories are fluorescent, e.g. powder used to lubricate the inside of plastic gloves. These may affect results if they contact optical surfaces. Using powder-free gloves removes a source of fluorescent contamination, as does touching reaction plates and caps only when necessary.
- Correct plastics. Use optical-grade reactions plates, tubes and caps, as these undergo special testing to ensure the absence of fluorescent impurities. Bent, creased or damaged plastics may adversely affect fluorescent signal transmission or prevent proper sealing of the well. This can result in evaporation, change in sample volume, and altered PCR chemistry. Inspect visually for damage before placing reaction plates in the thermal cycler. Be cautious about using plastics provided from vendors different from the instrument manufacturer, as they may not fit the thermal block properly and provide poor thermal transfer.
Low copy DNA target
When dispensing DNA samples (genomic DNA, plasmid, cDNA, etc.), if the average number of target molecules delivered per well is approximately 10 or lower, higher variability in the results is expected. This variability should conform to the Poisson distribution. For example, if an average of one target DNA molecule is dispensed per well, the Poisson distribution predicts 37% of the wells will receive zero target copies, 37% will receive 1 copy, 18% will receive 2 copies, 6% will receive 3 copies, etc.
One approach to the low copy DNA target problem is to increase the number of target DNA molecules dispensed per well to more than 10. This increase could be accomplished by increasing the sample volume delivered or using more concentrated samples. Another approach is to increase the number of replicates, which reduces the impact of low copy variability on the mean.