PCR, MIQE Guidelines

BACKGROUND:

Currently, a lack of consensus exists on how best to perform and interpret quantitative real- time PCR (qPCR) experiments. The problem is exac- erbated by a lack of sufficient experimental detail in many publications, which impedes a reader’s ability to evaluate critically the quality of the results presented or to repeat the experiments.

CONTENT:

The Minimum Information for Publication of Quantitative Real-Time PCR Experiments (MIQE) guidelines target the reliability of results to help ensure the integrity of the scientific literature, promote con- sistency between laboratories, and increase experimen- tal transparency. MIQE is a set of guidelines that de- scribe the minimum information necessary for evaluating qPCR experiments. Included is a checklist to accompany the initial submission of a manuscript to the publisher. By providing all relevant experimental conditions and assay characteristics, reviewers can as- sess the validity of the protocols used. Full disclosure of all reagents, sequences, and analysis methods is neces- sary to enable other investigators to reproduce results. MIQE details should be published either in abbreviated form or as an online supplement.

SUMMARY: 

Following these guidelines will encourage better experimental practice, allowing more reliable and unequivocal interpretation of qPCR results.

The fluorescence-based quantitative real-time PCR (qPCR)15(1–3 ）, with its capacity to detect and mea- sure minute amounts of nucleic acids in a wide range of samples from numerous sources, is the enabling tech- nology par excellence of molecular diagnostics, life sci- ences, agriculture, and medicine(4, 5 ). Its conceptual and practical simplicity, together with its combination of speed, sensitivity, and specificity in a homogeneous assay, have made it the touchstone for nucleic acid quantification. In addition to its use as a research tool, many diagnostic applications have been developed, in- cluding microbial quantification, gene dosage determi- nation, identification of transgenes in genetically mod- ified foods, risk assessment of cancer recurrence, and applications for forensic use (6 –11 ）.

This popularity is reflected in the prodigious number of publications reporting qPCR data, which invariably use diverse reagents, protocols, analysis methods, and reporting formats. This remarkable lack of consensus on how best to perform qPCR experiments has the adverse consequence of perpetuating a string of serious shortcomings that encumber its status as an independent yardstick(12）. Techni- cal deficiencies that affect assay performance include the following: (a) inadequate sample storage, prep- aration, and nucleic acid quality, yielding highly variable results; (b) poor choice of reverse-transcription primers and primers and probes for the PCR, leading to inefficient  and  less-than-robust  assay  performance;and (c) inappropriate data and statistical analyses, gen- erating results that can be highly misleading. Conse- quently, there is the real danger of the scientific litera- ture being corrupted with a multitude of publications reporting inadequate and conflicting results ( 13 ). The publication (14 ) and retraction (15 ) of a Science “Breakthrough of the Year 2005” report provides a disquieting warning. The problem is exacerbated by the lack of information that characterizes most reports of studies that have used this technology, with many publications not providing sufficient experimental detail to permit the reader to critically evaluate the quality of the results presented or to repeat the experi- ments. Specifically, information about sample acquisi- tion and handling, RNA quality and integrity, reverse- transcription details, PCR efficiencies, and analysis parameters are frequently omitted, whereas sample normalization is habitually carried out against single reference genes without adequate justification.

The aim of this document is to provide authors, reviewers, and editors specifications for the minimum information, set out in Table 1, that must be reported for a qPCR experiment to ensure its relevance, accu- racy, correct interpretation, and repeatability. MIQE (Minimum Information for Publication of Quantita- tive Real-Time PCR Experiments, pronounced mykee) is modeled on similar guidelines drawn up for DNA microarray analysis (16 ), proteomics experiments (17 ), genome sequence specification (18 ), and those under discussion for RNA interference work (19, 20 ) and metabolomics (21 ), all of which are initiatives co- ordinated under the umbrella of MIBBI (Minimum Information for Biological and Biomedical Investiga- tions, http://www.mibbi.org) (22 ). Compulsory inclu- sion of a common reporting language to allow data sharing is not proposed, although it is envisaged that a future update of these guidelines could include such a recommendation. Rather, these guidelines target the reliability of results to help ensure the integrity of the scientific literature, promote consistency between lab- oratories, and increase experimental transparency. They should be read in conjunction with recent publi- cations that deal in depth with the issue of qPCR stan- dardization (23–26 ).

1.Nomenclature

A few terms require standardization to ensure clarification:

1.1We propose that the abbreviation qPCR be used for quantitative real-time PCR and that RT-qPCR be used for reverse transcription– qPCR. Applying the abbreviation RT-PCR to qPCR causes confusion and is inconsistent with its use for conven- tional (legacy) reverse transcription–PCR.

1.2Genes used for normalization should be referred to as reference genes, not as housekeeping genes.

1.3TaqMan probes should be referred to as hydrolysis probes.

1.4The term FRET probe (fluorescence resonance en- ergy transfer probe) refers to a generic mechanism in which emission/quenching relies on the interac- tion between the electron-excitation states of 2 flu- orescent dye molecules. LightCycler-type probes should be referred to as dual hybridization probes.

1.5The Oxford English Dictionary lists only quantifica- tion, not quantitation; therefore, the former is the proper word.

1.6The nomenclature describing the fractional PCR cycle used for quantification is inconsistent, with threshold cycle (Ct), crossing point (Cp), and take-off point (TOP) currently used in the literature. These terms all refer to the same value from the real-time instrument and were coined by competing manu- facturers of real-time instruments for reasons of product differentiation, not scientific accuracy or clarity. We propose the use of quantification cycle (Cq), according to the RDML (Real-Time PCR Data Markup Language) data standard (http:// www.rdml.org) (27 ).

2.Conceptual Considerations

To explain and justify the guidelines, we find it useful to review a number of key issues surrounding qPCR experiments:

2.1Analytical sensitivity refers to the minimum num- ber of copies in a sample that can be measured accurately with an assay, whereas clinical sensitivity is the percentage of individuals with a given disor- der whom the assay identifies as positive for that condition. Typically, sensitivity is expressed as the limit of detection (LOD), which is the concentra- tion that can be detected with reasonable certainty (95% probability is commonly used) with a given analytical procedure. The most sensitive LOD the- oretically possible is 3 copies per PCR (28 ), assum- ing a Poisson distribution, a 95% chance of includ- ing at least 1 copy in the PCR, and single-copy detection. Experimental procedures typically in- clude sample-processing steps (i.e., extraction) and, when required, reverse transcription. If the volume changes and the efficiencies of these steps are accounted for, the most sensitive LOD theoret- ically possible can be expressed in units relevant to the experiment, such as copies per nanogram of tissue. Experimental results less than the theoretically possible LOD should never be reported. It also follows that results of “0” are meaningless and misleading. LOD estimates in qPCR analyses are complicated by the logarithmic nature of Cq, be- cause Cq is undefined when the template concen- tration is zero. Appropriate determination and modeling of the LOD in the qPCR is the focus of continued research (26 ).

2.2Analytical specificity refers to the qPCR assay de- tecting the appropriate target sequence rather than other, nonspecific targets also present in a sample. Diagnostic specificity is the percentage of individu- als without a given condition whom the assay iden- tifies as negative for that condition.

2.3Accuracy refers to the difference between experi- mentally measured and actual concentrations, presented as fold changes or copy number estimates.

2.4Repeatability (short-term precision or intraassay variance) refers to the precision and robustness of the assay with the same samples repeatedly ana- lyzed in the same assay. It may be expressed as the SD for the Cq variance. Alternatively, the SD or the CV for copy number or concentration variance may be used. CVs should not be used with Cqs, however (29 ).

2.5Reproducibility (long-term precision or interas- say variance) refers to the variation in results between runs or between different laboratories and is typically expressed as the SD or CV of copy numbers or concentrations. Cq values generated from different runs are subject to inherent interrun variation (30 ); hence, reporting interrun Cq variation is not appropriate.

Publications describing mRNA concentrations for target genes should make it clear precisely what the targets are. The transcripts of most human genes and many genes in other multicellular organisms are alter- natively spliced (31, 32 ), and these splicing variants specify alternative protein isoforms, with variation in splicing patterns reported in different tissues or at dif- ferent developmental stages. Consequently, single exon– based RT-qPCR assays may detect a number of splice variants, whereas intron-spanning primers may be more selective but may miss some splice variants altogether. Most recently, autosomal nonimprinted genes that display allelic imbalance in their expression have been described (33 ). Taken together, these find- ings imply that use of an RT-qPCR assay that simply targets one or at most 2 exons of an mRNA is no longer sufficient to describe the expression level of a particular gene. Consequently, sequence information for primers must be provided together with an assessment of their specificity with respect to known splice variants and single-nucleotide polymorphism positions docu- mented in transcript and single-nucleotide polymor- phism databases. For primer sets selected from the RTprimerDB database (34, 35 ), this is easily done by consulting the RTprimerDB Web site (http://www. rtprimerdb.org), which contains all the relevant infor- mation. For commercial assays, lot information and the providers’ experimental validation criteria are re- quired. The reporting of results for nonvalidated com- mercial assays and assays that have been validated only in silico are strongly discouraged.

It must be remembered that detection of the pres- ence of an mRNA provides no information on whether that mRNA will be translated into a protein or, indeed, whether a functional protein is translated at all.

Immunohistochemistry, western blotting,  or other protein-quantification methods are not always able to corroborate quantitative cellular mRNA data. It is now well established that there is frequently a lack of concordance between mRNA- and protein- concentration data (36 ), which is particularly true for mRNAs that specify proteins that are part of multi- function protein complexes (37 ). Finally, it has be- come clear that knowledge of the presence and func- tion of specific microRNAs is as important to understanding gene expression as being able to quan- tify the mRNA species (38 ).

It is also necessary to be aware that most quantita- tive RNA data are not absolute, but relative. Thus, the reference genes or materials used for standardization are critical, and any assessment of the validity of an RT-qPCR experiment must also consider the appropri- ateness of the relative-quantification reference. There- fore, the development of universal reference DNA and RNA calibration materials, although very  helpful  (39, 40 ), will not be a universal panacea (41, 42 ).

Much of the variance in reported expression values produced in RT-qPCR experiments is not simply due to variation in experimental protocols but is caused by corrections applied by various data- processing algorithms, each of which makes its own assumptions about the data. Consequently, although qPCR has frequently been proclaimed a touchstone or a gold standard, in practice this “standard” is a variable one, and the reporting of results requires considerable sophistication of analysis and interpretation (43 ).

3.Research vs Diagnostic Applications

Applications of qPCR technology can be broadly divided into research and diagnostic applications. Re- search applications usually analyze a wide range of targets with a fairly low throughput and many dif- ferent sample types. The main parameters that need to be addressed relate to assay analytical sensitivity and specificity, which in this context refer to how many target copies the assay can detect and whether the no-template controls (NTCs) are reliably nega- tive, respectively.

In contrast, diagnostic applications usually ana- lyze a limited number of targets, but require high-  throughput protocols that are targeted at only a few sample types. Although all of the considerations that apply to research applications also apply to diagnostic assays, clinical-diagnostic assays have a number of ad- ditional requirements that need to be considered. These requirements include information on analytical sensitivity and specificity that in this context refers to how often the assay returns a positive result when a target is present and how often it is negative in the absence of the target. Furthermore, the accuracy and precision within and between laboratories is often monitored by external QC programs. Additional clin- ical laboratory requirements include criteria for gener- ating reportable results, whether repeated measure- ments are made on samples, data on the resolution of false-positive/false-negative data, and the similarity of results from multiple laboratories that use the same and different technologies. Thus far, only a couple of interlaboratory comparisons have been performed, and both of these studies emphasized the need for stan- dardization of qPCR diagnostic assays (44, 45 ). An- other interlaboratory exercise is planned within the Eu- ropean Framework 7 project: SPIDIA (Standardisation and Improvement of Generic Pre-analytical Tools and Procedures for In-Vitro Diagnostics; http://www. spidia.eu).

4.Sample Acquisition, Handling, and Preparation

Sample acquisition constitutes the first potential source of experimental variability, especially for exper- iments targeting RNA, because mRNA profiles are eas- ily perturbed by sample-collection and -processing methods. There is some suggestion that fresh tissue can be stored on ice without major effects on RNA quality and concentration (46 ), but although this supposition may be true for some mRNAs and tissues, it may not be universally applicable. Hence, it is better to be cautious. Consequently, it is important to report in detail where the tissue sample was obtained and whether it was pro- cessed immediately. If the sample was not processed immediately, it is necessary to report how it was pre- served and how long and under what conditions it was stored.

A brief description of the sample is also essential.For example, microscopical examination of a tumor biopsy will reveal what percentage of the biopsy is made up of tumor cells, and this information should be reported.

Nucleic acid extraction is a second critical step. Extraction efficiency depends on adequate homogeni- zation, the type of sample (e.g., in situ tissue vs log- phase cultured cells), target density, physiological sta- tus (e.g., healthy, cancerous, or necrotic), genetic complexity, and the amount of biomass processed. Therefore, it necessary that details of the nucleic acid– extraction method be provided and that the methods used for measuring nucleic acid concentration and as- sessing its quality be described. Such details are partic- ularly crucial for RNA extracted from fresh frozen laser-microdissected biopsy samples, because varia- tions in tissue-preparation procedures have a substan- tial effect on both RNA yield and quality (47 ).

5.QC of Nucleic Acids

5.1.RNA SAMPLES

Quantification of RNA in the extracted samples is im- portant, because it is advisable that approximately the same amounts of RNA be used when comparing differ- ent samples. There are several quantification proce- dures in common use, however, including spectropho- tometry (NanoDrop; Thermo Scientific), microfluidic analysis (Agilent Technologies’ Bioanalyzer, Bio-Rad Laboratories’ Experion), capillary gel electrophoresis (Qiagen’s QIAxcel), or fluorescent dye detection (Am- bion/Applied Biosystems’ RiboGreen). The methods produce different results, making it unwise to try to compare data obtained with the different  methods (48 ). The preferred method for quantifying RNA uses fluorescent RNA-binding dyes (e.g., RiboGreen), which are best for detecting low target concentrations. In any case, it is advisable to measure all samples with a single method only and to report this information.

It is also important to test for and report the extent of genomic-DNA contamination and to record the threshold cutoff criteria for the amounts of such con- tamination that are tolerable. It is essential to report whether the RNA sample has been treated with DNase (including the type of DNase used and the reaction conditions) and to report the results from a compari-

son of Cqs obtained with positive and no–reverse tran- scription controls for each nucleic acid target.

It is also essential to document the quality assess- ment of RNA templates. The only situation in which this requirement does not apply is when the quantity of total RNA extracted is too low to permit quality assess- ment. This situation arises when RNA is extracted from single cells, plasma, other cell-free body fluids, some laser-captured samples, or clarified tissue culture me- dium. It also applies in cases in which extraction and RT-qPCR steps are performed as a continuous, single- tube experiment. Key information to report includes data on RNA quantity, integrity, and the absence of reverse transcription or PCR inhibitors. It is worth re- membering that RNA degrades markedly in vivo, ow- ing to the natural regulation of mRNAs in response to environmental stimuli (49 ). This source of RNA deg- radation is beyond the control of the researcher; one of its manifestations is that even high-quality RNA sam- ples can show differential degradation of individual mRNAs.

The A260/A280 ratio must be measured in a buffer at neutral pH, but such measurement is not sufficient if the nucleic acid is to be used for quantitative analysis, especially when the aim is to measure minor differ- ences (<10-fold) in cellular mRNA concentrations. The absorbance ratio does provide an indication of RNA purity, because the presence of DNA or residual phenol alters the ratio. Instead, one should provide gel electrophoresis evidence at the least or, better yet, re- sults from a microfluidics-based rRNA analysis (50 ) or a reference gene/target gene 3':5' integrity assay (51 ). The advantage of the use of a Bioanalyzer/Experion system to calculate an RNA integrity number or an RNA quality indicator number is that these measures provide quantitative information about the general state of the RNA sample. It is important to bear in mind, however, that these numbers relate to rRNA quality and cannot be expected to be an absolute mea- sure of quality. Use of a 3':5' assay requires that the PCR efficiencies of both assays be virtually identical (51 ) and not be subject to differential inhibition. This assay also necessitates the establishment of a threshold criterion that delineates the RNA quality sufficient to yield reliable results. Ideally, the assay should target a panel of “integrity reference genes,” probably without introns, with a 3':5' threshold ratio of approximately 0.2–5. Clearly, further work is required to generate a universally applicable, cost-effective, and simple pro- tocol for assessing RNA integrity.

Inhibition of reverse-transcription activity or PCR should be checked by dilution of the sample (prefera- bly) or use of a universal inhibition assay such as SPUD (52, 53 ). If the RNA sample is shown to be partially degraded, it is essential that this information be re- ported, because the assay’s sensitivity for detecting a low-level transcript may be reduced and relative differ- ences in the degradation of transcripts may produce incorrect target ratios.

5.2.DNA SAMPLES

In general, degradation is much less of an issue with DNA; however, it is important to be able to assess the extent of DNA degradation for forensic applications, i.e., in cases in which harsh environmental conditions at scenes of crimes or mass disasters or at sites involving missing-person cases may have degraded the chemical structure of DNA. The small amplicon size of qPCR assays helps to minimize assay-related problems, but methods have been developed that provide a quantita- tive measurement of DNA quality (54 ) and should be considered for such specialized purposes.

The potential for inhibition is a more generally applicable variable that must be addressed in a publi- cation, and it is important to ensure that no inhibitors copurified with the DNA will distort results, e.g., pathogen detection and their quantification (55 ). Al- though such approaches such as spiking samples with positive controls (52 ) can be used to detect inhibition, different PCR reactions may not be equally susceptible to inhibition by substances copurified in nucleic acid extracts (56, 57 ). Consequently, it is better to routinely use dilutions of nucleic acids to demonstrate that ob- served decreases in Cqs or copy numbers are consistent with the anticipated result and to report these data.

6.Reverse Transcription

The reverse-transcription step introduces substantial variation into an RT-qPCR assay (58, 59 ). Hence, it is essential that a detailed description of the protocol and reagents used to convert RNA into cDNA be provided. This documentation must include the amount of RNA reverse-transcribed, priming strategy, enzyme type, volume, temperature, and duration of the reverse- transcription step. It is recommended that the reverse- transcription step be carried out in duplicate or tripli- cate and that the total RNA concentration be the same in every sample (58 ).

7.qPCR

The following information must be provided for qPCR assays: database accession numbers of each target and reference gene, the exon locations of each primer and any probe, the sequences and concentra- tions of each oligonucleotide, including the identi- ties, positions, and linkages of any dyes and/or mod- ified bases. Also required are the concentration and identity of the polymerase, the amount of template (DNA or cDNA) in each reaction, the Mg2+ concen- tration, the exact chemical compositions of the buffer (salts, pH, additives), and the reaction vol- ume. The investigators must also identify the instru- ment they used and document all of the PCR cycling conditions. Because the consumables used affect thermal cycling, it is necessary to identify the use of single tubes, strips, or plates, and their manufactur- ers. The degree of transparency of the plasticware used, e.g., white or clear, is also important, because different plastics exhibit substantial differences in fluorescence reflection and sensitivity (60 ). When plates are used, the method of sealing (heat bonding vs adhesives) can affect the evaporation of samples at the plate perimeter and should therefore be documented.

Because PCR efficiency is highly dependent on the primers used, their sequences must be published. This requirement is perfectly feasible even with com- mercial primers, because there is a precedent for companies to make their primer and probe se- quences available (http://www.primerdesign.co.uk/ research_with_integrity.asp).

In addition, submission to public  databases such as RTprimerDB is strongly encouraged; over time, these databases could become universal clearinghouses.

7.1.SECONDARY STRUCTURE

The structure of the nucleic acid target (e.g., stem and loop secondary RNA structure) has a substantial im- pact on the efficiency of reverse transcription and the PCR. Therefore, the positions of primers, probes, and PCR amplicons must take the folding of RNA tem- plates into consideration. Sequences should be checked with nucleic acid–folding software, e.g., mfold for DNA (http://mfold.bioinfo.rpi.edu/cgi-bin/dna-form1.cgi) or RNA (http://frontend.bioinfo.rpi.edu/applications/ mfold/cgi-bin/rna-form1-2.3.cgi). Ideally, the folding structures should be made available to reviewers.

7.2.SPECIFICITY

In silico tools such as BLAST or equivalent specificity searches are useful for assay design. Any appreciable homology to pseudogenes or other unexpected targets should be documented and provided as aligned se- quences for review; however, specificity must be vali- dated empirically with direct experimental evidence (electrophoresis gel, melting profile, DNA sequencing, amplicon size, and/or restriction enzyme digestion).

Algorithms for predicting an oligonucleotide’s melting temperature (Tm) are useful for initial design, but the practical optimum temperature for annealing must be determined experimentally. Although primer optimization has become unfashionable, it is clear that poor annealing optimization has a large effect on assay quality (51 ). A marked presence of primer dimers pro- duces a lower PCR efficiency in probe-based assays and may generate false positives in assays based on SYBR Green I. Some evidence for primer optimization should be provided to reviewers, ideally in the form of annealing temperature or Mg2+ gradients, and be pre- sented as Cq values, plots of fluorescence vs cycle num- ber, and/or melting curves (61 ).

7.3.CONTROLS AND QUANTIFICATION CALIBRATORS

In addition to the no–reverse transcription control in RT-qPCR assays mentioned above, additional controls and/or quantification calibrators are required for all qPCR reactions. NTCs detect PCR contamination when probes are used and can also distinguish unin- tended amplification products (e.g., primer dimers) from the intended PCR products in SYBR Green I re- actions. NTCs should be included on each plate or batch of samples, and conditions for data rejection be

established. For example, NTCs with Cqs 2:40 could be ignored if the Cq for the lowest concentration un- known is 35.

Positive controls in the form of nucleic acids ex- tracted from experimental samples are useful for mon- itoring assay variation over time and are essential when calibration curves are not performed in each run.

Quantification calibrators may be purified target molecules, such as synthetic RNA or DNA oligonucle- otides spanning the complete PCR amplicon, plasmid DNA constructs, cDNA cloned into plasmids, RNA transcribed in vitro, reference RNA pools, RNA or DNA from specific biological samples, or internation- ally recognized biological standards (as they become available). Dilutions should be carried out into defined concentrations of carrier tRNA (yeast or Escherichia coli at 10 –100 ng/µL). For detection of human patho-gens, calibrators can be diluted into negative control sample RNA or DNA, or they can be diluted into healthy human plasma, after which lysis may be carried out in the presence of carrier tRNA. Serial dilutions of a particular template can be prepared as stock solutions that resist several freeze–thaw cycles. A fresh batch should be prepared when a Cq shift of 0.5–1.0 is de- tected. Alternatively, solutions for calibration curves can be stored for a week at 4 °C and then discarded.

For diagnostic assays, the qPCR should include an independently verified calibrator, if available, that lies within the linear interval of the assay. Positive and neg- ative extraction controls are also recommended.

7.4.ASSAY PERFORMANCE

The following assay performance characteristics must be determined: PCR efficiency, linear dynamic range, LOD, and precision.

7.4.1.PCR efficiency. Robust and precise qPCR assays are usually correlated with high PCR efficiency. PCR efficiency is particularly important when reporting mRNA concentrations for target genes relative to those of reference genes. The LlLlCq method is one of the most popular means of determining differences in con- centrations between samples and is based on normal- ization with a single reference gene. The difference in Cq values (LlCq) between the target gene and the refer- ence gene is calculated, and the LlCqs of the different samples are compared directly. The 2 genes must be amplified with comparable efficiencies for this com-parison to be accurate. The most popular method is not necessarily the most appropriate, however, and alter- native, more generalized quantitative models have been developed to correct for differences in amplifica- tion efficiency (62 ) and to allow the use of multiple reference genes (30 ).

PCR amplification efficiency must be established by means of calibration curves, because such calibra- tion provides a simple, rapid, and reproducible indica- tion of the mean PCR efficiency, the analytical sensitiv- ity, and the robustness of the assay. Amplification efficiency should be determined from the slope of the log-linear portion of the calibration curve. Specifically, PCR efficiency = 10—1/slope — 1, when the logarithm of the initial template concentration (the independent variable) is plotted on the x axis and Cq (the dependent variable) is plotted on the y axis. The theoretical max- imum of 1.00 (or 100%) indicates that the amount of product doubles with each cycle. Ideally, the CIs or SEs of the means of estimated PCR efficiencies are reported from replicated calibration curves.

Calibration curves for each quantified target must be included with the submitted manuscript so that this information can be made available to the reviewers; slopes and y intercepts derived from these calibration curves must be included with the publication. Differ- ences in PCR efficiency will produce calibration curves with different slopes. As a consequence, differences be- tween the Cq values of the targets and the references will not remain constant as template amounts are var- ied, and calculations of relative concentrations will be inaccurate, yielding misleading results.

Cq values >40 are suspect because of the implied low efficiency and generally should not be reported; however, the use of such arbitrary Cq cutoffs is not ideal, because they may be either too low (eliminating valid results) or too high (increasing false-positive re- sults) (26 ).

7.4.2.Linear dynamic range. The dynamic range over which a reaction is linear (the highest to the lowest quantifiable copy number established by means of a calibration curve) must be described. Depending on the template used for generating calibration curves, the dynamic range should cover at least 3 orders of magni- tude and ideally should extend to 5 or 6 log10 concen- trations. The calibration curve’s linear interval must include the interval for the target nucleic acids being quantified. Because lower limits of quantification are usually poorly defined, the variation at the lowest con- centration claimed to be within the linear interval should be determined. Correlation coefficients (r2 values) must be reported, and, ideally, CIs should be provided through the entire linear dynamic range.

 7.4.3.LOD. The LOD is defined as the lowest concen- tration at which 95% of the positive samples are de- tected. In other words, within a group of replicates con- taining the target at concentrations at the LOD, no more than 5% failed reactions should occur. Low-copy PCRs are stochastically limited, and LODs of <3 copies per PCR are not possible. If multiple reactions are per- formed, however, accurate quantification of lower concentrations can be  obtained  via  digital  PCR  (29, 63, 64 ). Indeed, concentration calibrators can be checked by limiting dilution to show that the percent- age of failed and successful reactions follows a Poisson distribution.

7.4.4.Precision. There are many explanations for vari- ation in qPCR results, including temperature differ- ences affecting the completion of annealing and/or de- naturation, concentration differences introduced by pipetting errors, and stochastic variation. Precision in qPCR typically varies with concentration, decreasing with the copy number. Ideally, intraassay variation (re- peatability) should be displayed in figures as SD error bars or as CIs on calibration curves with replicate sam- ples. CVs should not be used with Cqs (29 ) but can be used to express the variance in copy numbers or con- centrations. This technical variation should be distin- guished from biological variation. Biological replicates can directly address the statistical significance of differ- ences in qPCR results between groups or treatments. For diagnostic assays, it may also be necessary to report interassay precision (reproducibility) between sites and different operators.

7.5.MULTIPLEX qPCR

The ability to multiplex greatly expands the power of qPCR analysis (65, 66 ), particularly when applied to the simultaneous detection of point mutations or poly- morphisms (67 ). Multiplexing requires the presenta- tion of evidence demonstrating that accurate quantifi- cation of multiple targets in a single tube is not impaired, i.e., that assay efficiency and the LOD are the same as when the assays are run in uniplex fashion. This concern is of particular importance when targets of appreciably lower abundance are coamplified with highly abundant targets.

8.Data Analysis

Data analysis includes an examination of the raw data, an evaluation of their quality and reliability, and the generation of reportable results. Various data- collection and -processing strategies have been de- scribed, and a systematic evaluation has revealed that qPCR data-analysis methods differ substantially in their performance (68 ).

Detailed information on the methods of data anal- ysis and confidence estimation is necessary, together with specification of the software used. The methods of identifying outliers and the disposition of such data must be specified. Documenting assay precision re- quires identification of the statistical methods used to evaluate variances (e.g., 95% CIs) and presentation of the corresponding concentrations or Cq values. Such information should include both repeatability and re- producibility data, if available. As discussed above, re- porting of CVs for Cqs is inappropriate (29 ), because Cqs will always be lower (and therefore potentially mis- leading) than CVs calculated for copy numbers. Infor- mation must be provided on the methods used for as- sessing accuracy, including the statistical significance of reported differences between groups.

8.1.NORMALIZATION

Normalization is an essential component of a reliable qPCR assay because this process controls for variations in extraction yield, reverse-transcription yield, and ef- ficiency of amplification, thus enabling comparisons of mRNA concentrations across different samples. The use of reference genes as internal controls is the most common method for normalizing cellular mRNA data; however, although the use of reference genes is com- monly accepted as the most appropriate normalization strategy (69 ), their utility must be experimentally vali- dated for particular tissues or cell types and specific experimental designs. Unfortunately, although there is an increased awareness of the importance of systematic validation and although the potentially highly mislead- ing effects of the use of inappropriate reference genes for normalization are widely known, these consider- ations also are still widely disregarded (70 ). Conse- quently, many molecular analyses still contain qPCR data that are poorly normalized.

Normalization involves reporting the ratios of the mRNA concentrations of the genes of interest to those of the reference genes. Reference gene mRNAs should be stably expressed, and their abundances should show strong correlation with the total amounts of mRNA present in the samples.

Normalization against a single reference gene is not acceptable unless the investigators present clear ev- idence for the reviewers that confirms its invariant ex- pression under the experimental conditions described. The optimal number and choice of reference genes must be experimentally determined and the method reported (71–73 ).

8.2.VARIABILITY

The inherent variability of biological systems may rival or exceed experimental differences between groups. This variation is often observed when many biological replicates are used to increase the statistical significance of the experiment. Although differences between bio- logical replicates may be large, sufficient numbers may allow smaller experimental differences to be discerned. A recent publication provides a textbook example for handling such data and how to salvage biologically meaningful data from assays subject to high biological variation (74 ). Many factors contribute to experimen- tal variation and influence the number of biological replicates necessary to achieve a given statistical power. Consequently, power analysis is useful for determining the number of samples necessary for valid conclusions.

8.3.QUALITATIVE ANALYSIS

The use of the PCR to detect merely the presence of a nucleic acid template, rather than to quantify it accu- rately, is referred to as qualitative PCR, which is widely used in pathogen diagnostics. The problem with qual- itative/quantitative stratification of PCR methods is that an accurate yes/no answer requires information about the low-end sensitivity of the PCR assay. Conse- quently, even a qualitative assay should provide infor- mation about the assay’s performance characteristics, especially with respect to the points discussed in sec- tions 7.4.2. and 7.4.3.

Conclusions

The necessity for ensuring quality-assurance measures for qPCR and RT-qPCR assays is well recognized (25, 44, 75– 86 ). The main difference between qPCR and conventional (legacy) PCR assays is the expecta- tion of the former’s potential to quantify target nucleic acids accurately. This difference must be clearly recog- nized, and one cannot assume that legacy PCR assays can translate directly into the qPCR format. Table 1 provides a checklist for authors preparing a report of a qPCR study. Items deemed essential (E) are required to allow reviewers to evaluate the work and other investi- gators to reproduce it. Items considered desirable (D) are also important and should be included if possible, but they may not be available in all cases. Certainly, it is important to apply common sense: Compliance with all items on the checklist is not necessary for initial screening of expression signatures targeting hundreds of targets. Once a more limited set of targets (fewer than 20) has been identified, however, assay perfor- mance should be described as detailed by the checklist, which is hosted on http://www.rdml.org/miqe/.

In summary, the purpose of these guidelines is3-fold:

1.To enable authors to design and report qPCR exper- iments that have greater inherent value.

2.To allow reviewers and editors to measure the tech- nical quality of submitted manuscripts against an es- tablished yardstick.

3.To facilitate easier replication of experiments de- scribed in published studies that follow these guidelines.

As a consequence, investigations that use this widely applied technology will produce data that are more uniform, more comparable, and, ultimately, more reliable.

References

1.Higuchi R, Dollinger G, Walsh PS, Griffith R. Si- multaneous amplification and detection of spe- cific DNA sequences. Biotechnology (N Y) 1992; 10:413–7.

2.Higuchi R, Fockler C, Dollinger G, Watson R. Kinetic PCR analysis: real-time monitoring of DNA amplification reactions. Biotechnology (N Y) 1993;11:1026 –30.

3.Wittwer CT, Herrmann MG, Moss AA, Rasmussen RP. Continuous fluorescence monitoring of rapid cycle DNA amplification. Biotechniques 1997;22: 130–8.

4.Bustin SA. Absolute quantification of mRNA using real-time reverse transcription polymer- ase chain reaction assays. J Mol Endocrinol 2000;25:169 –93.

5.Kubista M, Andrade JM, Bengtsson M, Forootan A, Jonak J, Lind K, et al. The real-time polymerase chain reaction. Mol Aspects Med 2006;27:95– 125.

6.Bernard PS, Wittwer CT. Real-time PCR technol- ogy for cancer diagnostics. Clin Chem 2002;48: 1178 – 85.

7.Mackay IM, Arden KE, Nitsche A. Real-time PCR in virology. Nucleic Acids Res 2002;30:1292–305.

8.Mackay IM. Real-time PCR in the microbiology laboratory. Clin Microbiol Infect 2004;10:190 – 212.

9.Bustin SA, Mueller R. Real-time reverse tran- scription PCR (qRT-PCR) and its potential use in clinical diagnosis. Clin Sci (Lond) 2005;109: 365–79.

10.Bustin SA, Mueller R. Real-time reverse transcrip- tion PCR and the detection of occult disease in colorectal cancer. Mol Aspects Med 2006;27: 192–223.

11.van den Berg RJ, Vaessen N, Endtz HP, Schulin T, van der Vorm ER, Kuijper EJ. Evaluation of real- time PCR and conventional diagnostic methods for the detection of Clostridium difficile- associated diarrhoea in a prospective multicentre study. J Med Microbiol 2007;56:36 – 42.

12.Bustin SA, Nolan T. Pitfalls of quantitative real- time reverse-transcription polymerase chain reac- tion. J Biomol Tech 2004;15:155– 66.

13.Garson JA, Huggett JF, Bustin SA, Pfaffl MW, Benes V, Vandesompele J, Shipley GL. Unreliable real-time PCR analysis of human endogenous retrovirus-W (HERV-W) RNA expression and DNA copy number in multiple sclerosis. AIDS Res Hum Retroviruses. Forthcoming 2009.

14.Huang T, Bohlenius H, Eriksson S, Parcy F, Nilsson O. The mRNA of the Arabidopsis gene FT moves from leaf to shoot apex and induces flowering. Science 2005;309:1694 – 6.

15.Bohlenius H, Eriksson S, Parcy F, Nilsson O. Retraction. Science 2007;316:367.

16.Brazma A, Hingamp P, Quackenbush J, Sher- lock G, Spellman P, Stoeckert C, et al. Mini- mum information about a microarray experiment (MIAME)—toward standards for microarray data. Nat Genet 2001;29:365–71.

17.Taylor CF, Paton NW, Lilley KS, Binz PA, Julian RK Jr, Jones AR, et al. The minimum information about a proteomics experiment (MIAPE). Nat Bio- technol 2007;25:887–93.

18.Field D, Garrity G, Gray T, Morrison N, Selengut J, Sterk P, et al. The minimum information about a genome sequence (MIGS) specification. Nat Bio- technol 2008;26:541–7.

19.Echeverri CJ, Beachy PA, Baum B, Boutros M, Buchholz F, Chanda SK, et al. Minimizing the risk of reporting false positives in large-scale RNAi screens. Nat Methods 2006;3:777–9.

20.Haney SA. Increasing the robustness and validity of RNAi screens. Pharmacogenomics 2007;8: 1037– 49.

21.Sansone SA, Fan T, Goodacre R, Griffin JL, Hardy NW, Kaddurah-Daouk R, et al. The metabolomics standards initiative. Nat Biotechnol 2007;25:846–8.

22.Taylor CF, Field D, Sansone SA, Aerts J, Apweiler R, Ashburner M, et al. Promoting coherent mini- mum reporting guidelines for biological and bio- medical investigations: the MIBBI project. Nat Biotechnol 2008;26:889 –96.

23.Burns MJ, Valdivia H, Harris N. Analysis and interpretation of data from real-time PCR trace detection methods using quantitation of GM soya as a model system. Anal Bioanal Chem 2004;378: 1616 –23.

24.Burns MJ, Nixon GJ, Foy CA, Harris N. Standar- disation of data from real-time quantitative PCR methods – evaluation of outliers and comparison of calibration curves. BMC Biotechnol 2005;5:31.

25.Ellison SL, English CA, Burns MJ, Keer JT. Routes to improving the reliability of low level DNA analysis using real-time PCR. BMC Biotechnol 2006;6:33.

26.Burns MJ, Valdivia H. Modelling the limit of de- tection in real-time quantitative PCR. Eur Food Res Technol 2008;226:1513–24.

27.Lefever S, Hellemans J, Pattyn F, Przybylski DR, Taylor C, Geurts R, et al. RDML: structured language and reporting guidelines for real-time quantitative PCR data. Nucleic Acids Res [Epub ahead of print 2009 Feb 17]. Available at: http://nar.oxfordjournals. org/cgi/content/abstract/gkp056.

28.Wittwer CT, Kusakawa N. Real-time PCR. In: Persing DH, Tenover FC, Versalovic J, Tang JW, Unger ER, Relman DA, White TJ, eds. Molecular microbiology: diagnostic principles and practice. Washington: ASM Press; 2004. p 71– 84.

29.Schmittgen TD, Livak KJ. Analyzing real-time PCR data by the comparative CT method. Nat Protoc 2008;3:1101– 8.

30.Hellemans J, Mortier G, De Paepe A, Speleman F, Vandesompele J. qBase relative quantification framework and software for management and automated analysis of real-time quantitative PCR data. Genome Biol 2007;8:R19. 31.de la Grange P, Dutertre M, Correa M, Auboeuf D. A new advance in alternative splicing databases: from catalogue to detailed analysis of regulation of expression and function of human alternative splicing variants. BMC Bioinformatics 2007;8:180. 32.Ben-Dov C, Hartmann B, Lundgren J, Valcarcel J. Genome-wide analysis of alternative pre-mRNA splicing. J Biol Chem 2008;283:1229 –33. 33.Bjornsson HT, Albert TJ, Ladd-Acosta CM, Green RD, Rongione MA, Middle CM, et al. SNP-specific array-based allele-specific expression analysis. Genome Res 2008;18:771–9. 34.Pattyn F, Robbrecht P, De Paepe A, Speleman F, Vandesompele J. RTPrimerDB: the real-time PCR primer and probe database, major update 2006. Nucleic Acids Res 2006;34:D684 – 8. 35.Pattyn F, Speleman F, De Paepe A, Vandesompele J.RTPrimerDB: the real-time PCR primer and probe database. Nucleic Acids Res 2003;31: 122–3. 36.Gygi SP, Rochon Y, Franza BR, Aebersold R. Correlation between protein and mRNA abun- dance in yeast. Mol Cell Biol 1999;19:1720 –30. 37.Schmidt MW, Houseman A, Ivanov AR, Wolf DA. Comparative proteomic and transcriptomic profil- ing of the fission yeast Schizosaccharomyces pombe. Mol Syst Biol 2007;3:79. 38.Landi D, Gemignani F, Naccarati A, Pardini B, Vodicka P, Vodickova L, et al. Polymorphisms within micro-RNA-binding sites and risk of spo- radic colorectal cancer. Carcinogenesis 2008;29: 579 – 84. 39.Cronin M, Ghosh K, Sistare F, Quackenbush J, Vilker V, O’Connell C. Universal RNA reference materials for gene expression. Clin Chem 2004; 50:1464 –71. 40.Novoradovskaya N, Whitfield ML, Basehore LS, Novoradovsky A, Pesich R, Usary J, et al. Univer- sal Reference RNA as a standard for microarray experiments. BMC Genomics 2004;5:20. 41.Gingeras TR. RNA reference materials for gene expression studies. Difficult first steps. Clin Chem 2004;50:1289 –90. 42.Joseph LJ. RNA reference materials for gene ex- pression studies. RNA metrology: forecast calls for partial clearing. Clin Chem 2004;50:1290 –2. 43.Bustin SA, Benes V, Nolan T, Pfaffl MW. Quanti- tative real-time RT-PCR—a perspective. J Mol Endocrinol 2005;34:597– 601. 44.Ramsden SC, Daly S, Geilenkeuser WJ, Duncan G, Hermitte F, Marubini E, et al. EQUAL-quant: an international external quality assessment scheme for real-time PCR. Clin Chem 2006;52:1584 –91. 45.Damond F, Benard A, Ruelle J, Alabi A, Kupfer B, Gomes P, et al. Quality control assessment of human immunodeficiency virus type 2 (HIV-2) viral load quantification assays: results from an international collaboration on HIV-2 infection in 2006. J Clin Microbiol 2008;46:2088 –91. 46.Micke P, Ohshima M, Tahmasebpoor S, Ren ZP, Ostman A, Ponten F, Botling J. Biobanking of fresh frozen tissue: RNA is stable in nonfixed surgical specimens. Lab Invest 2006;86:202–11. 47.Morrogh M, Olvera N, Bogomolniy F, Borgen PI, King TA. Tissue preparation for laser capture microdissection and RNA extraction from fresh frozen breast tissue. Biotechniques 2007;43: 41–2, 4, 6 passim. 48.Bustin SA. Real-time, fluorescence-based quanti- tative PCR: a snapshot of current procedures and preferences. Expert Rev Mol Diagn 2005;5: 493– 8. 49.Doma MK, Parker R. RNA quality control in eu- karyotes. Cell 2007;131:660 – 8.

50.Fleige S, Pfaffl MW. RNA integrity and the effect on the real-time qRT-PCR performance. Mol As- pects Med 2006;27:126 –39.

51.Nolan T, Hands RE, Bustin SA. Quantification of mRNA using real-time RT-PCR. Nat Protoc 2006; 1:1559 – 82.

52.Nolan T, Hands RE, Ogunkolade BW, Bustin SA. SPUD: a qPCR assay for the detection of inhibitors in nucleic acid preparations. Anal Biochem 2006; 351:308 –10.

53.Ferns RB, Garson JA. Development and evalua- tion of a real-time RT-PCR assay for quantifica- tion of cell-free human immunodeficiency virus type 2 using a brome mosaic virus internal con- trol. J Virol Methods 2006;135:102– 8.

54.Swango KL, Hudlow WR, Timken MD, Buoncris- tiani MR. Developmental validation of a multiplex qPCR assay for assessing the quantity and quality of nuclear DNA in forensic samples. Forensic Sci Int 2007;170:35– 45.

55.Garson JA, Grant PR, Ayliffe U, Ferns RB, Tedder RS. Real-time PCR quantitation of hepatitis B virus DNA using automated sample preparation and murine cytomegalovirus internal control. J Virol Methods 2005;126:207–13.

56.Huggett JF, Novak T, Garson JA, Green C, Morris- Jones SD, Miller RF, Zumla A. Differential suscep- tibility of PCR reactions to inhibitors: an impor- tant and unrecognised phenomenon. BMC Res Notes 2008;1:70.

57.Ståhlberg A, Aman P, Ridell B, Mostad P, Kubista M. Quantitative real-time PCR method for detec- tion of B-lymphocyte monoclonality by compari- son of K and 入 immunoglobulin light chain ex- pression. Clin Chem 2003;49:51–9.

58.Ståhlberg A, Hakansson J, Xian X, Semb H, Kubista M. Properties of the reverse transcription reaction in mRNA quantification. Clin Chem 2004;50:509 –15.

59.Ståhlberg A, Kubista M, Pfaffl M. Comparison of reverse transcriptases in gene expression analy- sis. Clin Chem 2004;50:1678 – 80.

60.Reiter M, Pfaffl M. Effects of plate position, plate type and sealing systems on real-time PCR re- sults. Biotechnol Biotechnol Equip 2008;22: 824–8.

61.Ririe KM, Rasmussen RP, Wittwer CT. Product differentiation by analysis of DNA melting curves during the polymerase chain reaction. Anal Bio- chem 1997;245:154 – 60.

62.Pfaffl MW. A new mathematical model for rela- tive quantification in real-time RT-PCR. Nucleic Acids Res 2001;29:E45.

63.Dube S, Qin J, Ramakrishnan R. Mathematical analysis of copy number variation in a DNA sam- ple using digital PCR on a nanofluidic device. PLoS ONE 2008;3:e2876.

64.Vogelstein B, Kinzler KW. Digital PCR. Proc Natl Acad Sci U S A 1999;96:9236 – 41.

65.Elnifro EM, Ashshi AM, Cooper RJ, Klapper PE. Multiplex PCR: optimization and application in diagnostic virology. Clin Microbiol Rev 2000;13: 559 –70.

66.Wittwer CT, Herrmann MG, Gundry CN, Elenitoba-Johnson KS. Real-time multiplex PCR assays. Methods 2001;25:430 – 42.

67.Elenitoba-Johnson KS, Bohling SD, Wittwer CT, King TC. Multiplex PCR by multicolor fluorimetry and fluorescence melting curve analysis. Nat Med 2001;7:249 –53.

68.Karlen Y, McNair A, Perseguers S, Mazza C, Mer- mod N. Statistical significance of quantitative PCR. BMC Bioinformatics 2007;8:131.

69.Huggett J, Dheda K, Bustin S, Zumla A. Real-time RT-PCR normalisation; strategies and consider- ations. Genes Immun 2005;6:279 – 84.

70.Gutierrez L, Mauriat M, Guenin S, Pelloux J, Lefebvre JF, Louvet R, et al. The lack of a system- atic validation of reference genes: a serious pitfall undervalued in reverse transcription-polymerase chain reaction (RT-PCR) analysis in plants. Plant Biotechnol J 2008;6:609 –18.

71.Vandesompele J, De Preter K, Pattyn F, Poppe B, Van Roy N, De Paepe A, Speleman F. Accu- rate normalization of real-time quantitative RT- PCR data by geometric averaging of multiple internal control genes. Genome Biol 2002;3: RESEARCH0034.

72.Pfaffl MW, Horgan GW, Dempfle L. Relative ex- pression software tool (REST) for group-wise comparison and statistical analysis of relative expression results in real-time PCR. Nucleic Acids Res 2002;30:e36.

73.Andersen CL, Jensen JL, Orntoft TF. Normalization of real-time quantitative reverse transcription- PCR data: a model-based variance estimation approach to identify genes suited for normaliza- tion, applied to bladder and colon cancer data sets. Cancer Res 2004;64:5245–50.

74.Willems E, Leyns L, Vandesompele J. Standard- ization of real-time PCR gene expression data from independent biological replicates. Anal Bio- chem 2008;379:127–9.

75.Apfalter P, Reischl U, Hammerschlag MR. In- house nucleic acid amplification assays in research: How much quality control is needed before one can rely upon the results? J Clin Microbiol 2005;43:5835– 41.

76.Ciabatti I, Froiio A, Gatto F, Amaddeo D, Marchesi U. In-house validation and quality con- trol of real-time PCR methods for GMO detection: a practical approach. Dev Biol (Basel) 2006;126: 79 – 86; discussion 324 –5.

77.de Cremoux P, Bieche I, Tran-Perennou C, Vig- naud S, Boudou E, Asselain B, et al. Inter-laboratory quality control for hormone-dependent gene ex- pression in human breast tumors using real-time reverse transcription-polymerase chain reaction. Endocrine Relat Cancer 2004;11:489 –95.

78.de Vries TJ, Fourkour A, Punt CJ, van de Locht LT, Wobbes T, van den Bosch S, et al. Reproducibility of detection of tyrosinase and MART-1 transcripts in the peripheral blood of melanoma patients: a quality control study using real-time quantitative RT-PCR. Br J Cancer 1999;80:883–91.

79.Gabert J, Beillard E, van der Velden VH, Bi W, Grimwade D, Pallisgaard N, et al. Standardization and quality control studies of ‘real-time’ quanti- tative reverse transcriptase polymerase chain re- action of fusion gene transcripts for residual dis- ease detection in leukemia – a Europe Against Cancer program. Leukemia 2003;17:2318 –57.

80.Lemmer K, Donoso Mantke O, Bae HG, Groen J, Drosten C, Niedrig M. External quality control assessment in PCR diagnostics of dengue virus infections. J Clin Virol 2004;30:291– 6.

81.Marubini E, Verderio P, Raggi CC, Pazzagli M, Orlando C. Statistical diagnostics emerging from external quality control of real-time PCR. Int J Biol Markers 2004;19:141– 6.

82.Raggi CC, Verderio P, Pazzagli M, Marubini E, Simi L, Pinzani P, et al. An Italian program of external quality control for quantitative assays based on real-time PCR with Taq-Man probes. Clin Chem Lab Med 2005;43:542– 8.

83.Sjoholm MI, Dillner J, Carlson J. Assessing quality and functionality of DNA from fresh and archival dried blood spots and recommendations for qual- ity control guidelines. Clin Chem 2007;53: 1401–7.

84.Wang X, Jia S, Meyer L, Xiang B, Chen LY, Jiang N, et al. Comprehensive quality control utiliz- ing the prehybridization third-dye image leads to accurate gene expression measurements by cDNA microarrays. BMC Bioinformatics 2006; 7:378.

85.Winters MA, Tan LB, Katzenstein DA, Merigan TC. Biological variation and quality control of plasma human immunodeficiency virus type 1 RNA quantitation by reverse transcriptase poly- merase chain reaction. J Clin Microbiol 1993;31: 2960 – 6.

86.van der Velden VH, Panzer-Grumayer ER, Cazza- niga G, Flohr T, Sutton R, Schrauder A, et al. Optimization of PCR-based minimal residual dis- ease diagnostics for childhood acute lymphoblas- tic leukemia in a multi-center setting. Leukemia 2007;21:706 –13.