Interlaboratory diagnostic validation of conformation-sensitive capillary electrophoresis for mutation scanning.
Prescreening strategies aimed at identifying variant-carrying PCR products without fully characterizing them are of considerable interest because they greatly reduce the requirement for confirmatory sequencing (1, 2). In recent years, the investigation of hereditary breast cancer has presented a substantial problem for genetic-diagnostics laboratories. Currently, a full screen consists of mutation scanning of the coding regions and flanking intronic sequences of the 2 hereditary breast cancer genes: BRCA1  (breast cancer 1, early onset; MIM 113705), which is located at 17q21 and consists of 5589 bp of coding sequence distributed over 22 exons (3), and BRCA2 (breast cancer 2, early onset; MIM 600185), which is at 13q12.3 and consists of 10 257 bp of coding sequence distributed over 26 exons (4). More than 3400 different BRCA1 and BRCA2 variants have been identified (5), with the overwhelming majority being restricted to only a few families. Apart from a few common mutations in specific populations, such as Ashkenazi Jews, variants are reasonably evenly distributed throughout the 2 genes, a fact that rules out the possibility of a focused screening approach.
Many different prescreening methodologies have been applied to breast cancer genes, including denaturing gradient gel electrophoresis (6, 7), single-strand conformation polymorphism analysis (8, 9), hetero-duplex analysis (10), conformation-sensitive gel electrophoresis (11,12), denaturing HPLC (13), and mismatch-cleavage assays (14). The throughput of these techniques is limited, however, and their diagnostic accuracies have generally been poorly reported in the literature (15).
More recently, high-resolution melting analysis (16-18) has been used. This technique allows a substantially higher throughput but requires intensive fragment-specific optimization (17). In the absence of appropriate validation, the detection of certain insertions and deletions of single base pairs in homopolymer stretches may be questionable (see Supplemental Data I in the Data Supplement that accompanies the online version of this article at http://www.clinchem. org/content/vol56/issue4).
Conformation-sensitive capillary electrophoresis (CSCE)  is a simple method for indirect mutation scanning that is based on the principle that homoduplex and heteroduplex DNA have different electrophoretic mobilities under partially denaturing conditions (19-21). Fluorescently labeled PCR products are denatured and reannealed, allowing heteroduplexes to form. Wild-type samples can form onlya single homoduplex species, whereas heterozygous samples can form 4 possible duplex species: 2 heteroduplexes and 2 homoduplexes. Theoretically, these 4 species have different conformations under partially denaturing conditions and migrate at different rates during electrophoresis. In practice, however, one or more of the different species often remain unresolved.
In common with most mutation-scanning methods, including Sanger sequencing, CSCE typically will not detect large deletions or duplications. A recently described approach that combines fluorescence multiplex quantitative PCR with CSCE does facilitate detection of these mutations (22); however, a diagnostic strategy for genes such as BRCA1 and BRCA2, for which such mutations are important, usually requires the inclusion of mutation-detection techniques such as multiplex ligation-dependent probe amplification or quantitative PCR.
[FIGURE 1 OMITTED]
We report optimized conditions for CSCE analysis, together with an extensive formal diagnostic validation with the breast cancer genes as a model.
Materials and Methods
The study was carried out in 3 laboratories (laboratories 1, 2, and 3). Laboratories 1 and 2 had previous experience with CSCE and performed the electrophoresis experiments on Applied Biosystems 3730 DNA Analyzer instruments (48 capillaries), whereas laboratory 3 had no prior experience and used an Applied Biosystems 3130x/ Genetic Analyzer (16 capillaries).
CSCE conditions were based on the method developed at the Sanger Institute (21). The separation matrix consisted of 50 g/L Conformation Analysis Polymer (Applied Biosystems) in 1X TTE Glycerol Tolerant Buffer (which contains Tris, taurine, and EDTA; National Diagnostics), with varying concentrations of urea and sucrose. Applied Biosystems supplied special run modules that allowed the run temperature to be reduced to 10 [degrees]C.
Manual analysis by visual inspection was carried out in GeneMapper software (version 3.7; Applied Biosystems). To determine the quantitative affect of varying technical parameters, we scored all analyses (except those indistinguishable from the wild-type control, which were scored 0) by assigning them a resolution score, which was equal to the number of peaks present plus shoulders, which counted as 0.5 (Fig. 1). Traces with peaks outside the analysis window for the instrument [500-6000 relative fluorescent units (RFU) for the 3130xl instrument and 1000-30 000 RFU for the 3730 instruments] were scored as failures.
The automated analysis used the heteroduplex analysis plug-in for BioNumerics software (version 5.10; Applied Maths) (23). Initially, traces with peaks below a defined threshold are excluded. Then, 5 parameters are used to compare the remaining traces with a specified wild-type control. The parameters used were SRMS (single-sided root mean square), which measures the percent peak broadening and compares it with that of the wild-type control, and MAXDIFF, which is largest vertical difference between 2 compared curves for the test and wild-type traces. The sum (SRMS + MAXDIFF) was used as a single measure of the efficacy of automated analysis. The other 3 parameters (SECPK, DFH3, and DFH4) were not used.
We used a systematic panel of 52 plasmid-based controls for the optimization phase (24). The physical parameters tested were the G + C content of the fragment, the position of the variant within the fragment, and the specific nature of the heteroduplex mismatch (single-base variants only). The process parameters tested were PCR dilution, electrophoresis loading parameters, urea content of the polymer, run temperature, and run voltage. In addition, we performed a series of experiments to determine the CSCE limit of detection with respect to the concentration of the variant allele in the sample. Full details of optimization experiments are described in Supplemental Data A in the online Data Supplement.
FORMAL BLINDED VALIDATION
The validation was performed with 79 PCR amplicons (235-509 bp in length) that covered the entire coding regions of the BRCA1 and BRCA2 genes (see Supplemental Data B in the online Data Supplement). The total analyzable region of interest was 23 769 bp. For validation we used blinded, anonymized patient samples that had previously tested positive for one or more BRCA1 or BRCA2 variants in one of the 3 laboratories. We selected 359 patient samples on the basis of (a) availability of DNA for analysis, (b) presence of nonredundant variants, and (c) exclusion of variants originally discovered by CSCE or a similar methodology.
We arrayed the DNA samples over 10 plates in 79 blocks that represented the 79 PCR amplicons analyzed (see Supplemental Data C in the online Data Supplement). Except for the 2 wild-type controls identified in each block, analyses within blocks were randomized and blinded. The final analysis panel consisted of 365 wild-type, 445 mutant, and 150 control analyses (n = 960). Because of the replication of variants, the mutant analyses consisted of 402 different variants: 267 point mutations, 24 insertions or duplications of up to 11 bp, 93 deletions of up to 62 bp, 2 insertion-deletions, and 16 compound variants (>1 variant in a fragment) (see Supplemental Data D in the online Data Supplement). The sample panel was confirmed by sequencing on a 3130xl Genetic Analyzer.
To facilitate fluorescence labeling, we used a 3-primer system in each laboratory to carry out the PCRs (25). For the purposes of manual scoring, 2 operators independently inspected all analyses. When the analysts' results differed, the more conservative result (i.e., with the lower variant score) was taken.
LOSS OF RESOLUTION
During the pilot analysis, we observed a cumulative loss in peak resolution with sequentially loaded plates in laboratory 1, but not in laboratory 2 or 3. This issue is discussed fully in Supplemental Data E in the online Data Supplement.
OPTIMIZATION OF CSCE RUN CONDITIONS
Full details of the optimization phase are described in Supplemental Data A in the online Data Supplement. Optimal conditions for the analyses of fragments with 20%-60% G + C content centered around 12.5 kV at 14 [degrees]C with 46 g/L urea, whereas the optimal conditions for fragments with 80% G + C content were 12.5 kV at 18 [degrees]C with 60 g/L urea (Fig. 2). Optimal cutoffs for the automated analysis parameters SRMS and MAXDIFF were 8.0 and 6.0, respectively.
LIMIT OF DETECTION
All but one of the mutations were detected at relative concentrations of [greater than or equal to] 22%. Below this percentage, however, the detection rate dropped to 0% for all variants at 3% representation (see Supplemental Data A in the online Data Supplement).
formal blinded validation
The panel of BRCA1 and BRCA2 samples was independently tested in each laboratory with the conditions determined in the optimization phase (validation raw data are given in Supplemental Data F in the online Data Supplement). The overall failure rate in the automated analysis was 5% (143 of 2880 analyses), ranging from 2.5% (24 of 960 analyses) in laboratory 1 to 7% (71 of 960 analyses) in laboratory 2. Both laboratory 1 and laboratory 2 detected all variants, but laboratory 3 failed to detect one of the variants. The overall false-positive rate was 8.3% (84 of 1011 analyses), ranging from 2% (7 of 358 analyses) in laboratory 1 to 13.4% (44 of 328 analyses) in laboratory 3.
[FIGURE 2 OMITTED]
In the manual analysis, the overall failure rate was 5% (142 of 2880 analyses), ranging from 1% (11 of 960 analyses) in laboratory 1 to 8% (77 of 960 analyses) in laboratory 3. The overall false-positive rate was 5.6% (57 of 1015 analyses), ranging from 1.6% (5 of 314 analyses) in laboratory 3 to 10.6% (36 of 339 analyses) in laboratory 2. All variants were detected in laboratory 1, but laboratories 2 and 3 missed 1 variant and 11 variants, respectively. Blinded reanalysis of the data for laboratory 3 showed that 5 of the missed variants were clearly detectable, with at least a shouldered peak to indicate their presence. The trace indicating a variant's presence was subtle for 1 variant. Three of these variants were indistinguishable from the control peak, but 2 of these variants should have been detected but were not because of poor amplification (see Supplemental Data G in the online Data Supplement).
[FIGURE 3 OMITTED]
The final variant not detected by manual analysis in laboratory 3 was BRCA1:c.4676-34G>A, which is upstream of exon 16 and is located 66 bp from the end of the amplicon (G + C content, 45.5%). This variant also evaded detection by automated analysis in laboratory 3 and was the single variant not detected by manual analysis in laboratory 2. It represents the subtlest electrophoretic shift seen in this validation (Fig. 3A).
All of the insertion and deletion variants and 93% of the point variants produced at least 2 distinct peaks (95% overall). Only 5 of the point variants (<2%) consistently yielded a resolution score of 1.
A marginal loss of detection efficacy was seen in fragments with a higher G + C content and with variants at the ends of the fragments, but there were no statistically significant correlations (Fig. 4).
The reproducibility of the variant peak shapes was measured by Spearman rank correlation analysis of resolution scores (for manual analysis) and SRMS + MAXDIFF analysis (for automated analysis). The distributions of variant analysis scores between laboratories were well correlated in terms of monotonic ordinality for both the manual and automated analyses (i.e., ifthe variants were placed in order of their analysis score within each laboratory, they would, in general, appear in the same order [ P < 0.0001 in each case, Spearman rank correlation; null hypothesis (H0): Lab x [not equal to] Lab y].
The distributions of variant analysis scores were not correlated with respect to location or dispersal (P < 0.0001, Friedman test; H0: Lab x = Lab y). For automated analysis, differences were caused by constant bias. With the results of laboratory 2 used as the baseline, laboratories 1 and 3 had a mean bias of -6.7 and -4.9, respectively; however, although laboratory 1 yielded the lowest scores for the variants in the automated analysis, manual analysis produced the highest mean resolution scores (3.1 for laboratory 1; 3.0 for laboratory 2; 2.9 for laboratory 3). In addition, the data for laboratory 1 were notably less widely distributed than the data for the other laboratories (SRMS + MAXDIFF SDs: 10.9 for laboratory 1; 16.9 for laboratory 2; 19.4 for laboratory 3).
[FIGURE 4 OMITTED]
SENSITIVITY AND SPECIFICITY
The accuracy of sequencing is such that any alternative must be validated to a very high standard if it is to be used for diagnostic purposes. In this study, we have shown that optimal performance for the analysis of fragments with 20%-60% G + C content can be achieved with a single set of run conditions (see Supplemental Data A in the online Data Supplement). Because only 23% of the exons in the genome (NCBI Build 36.3) have a G + C content >60%, the conditions we have validated should be applicable to most diagnostic situations.
Our extensive blinded validation shows that the sensitivity of CSCE for detecting heterozygous variants in fragments <500 bp with a G + C content of 20%-60% is 98.3%-99.4% (95% CI) for manual analysis and >99.6% (95% CI) for automated analysis (Table 1). At these high levels, it is likely that interlaboratory variation would have a more important effect on sensitivity than any inaccuracy in estimating sensitivity. Therefore, for practical purposes we have shown that the sensitivity of CSCE for detecting heterozygous variants is indistinguishable from that for sequencing.
The overall specificity is dependent on the method of analysis used. With manual analysis, the estimated specificity is 93.1%-95.5% (95% CI), whereas the estimated specificity with automated analysis is 90.1%-93.1% (95% CI). The lower specificity achieved with automated analysis is largely due to the low-level background noise, which is often distant from the actual sample peak. Such noise can easily be excluded by visual inspection but cannot be identified as such by the analysis program (see Supplemental Data H in the online Data Supplement). Although an increase in the false-positive rate with automated analysis increases the number of fragments requiring sequencing, the difference is marginal. The key benefit of automated analysis is its objectivity and consistency, which underpin a reliable sensitivity. Automated analysis also enables the development of schemes for objectively monitoring resolution and facilitates a substantial reduction in the time required for analysis.
A 50-bp buffer was included in the PCR designs to allow for decreased sensitivity in detecting variants close to the ends of fragments, an observation that has previously been reported (21). Seventeen different variants included in this study fell within the stipulated 50-bp buffer, with distances as close as 36 bp from the end of the amplified fragment. Although we detected all of these variants, we found a slight bias toward lower resolution scores in this group compared with the validation as a whole (Fig. 4). Therefore, a minimum 50-bp buffer is a prudent measure to maintain performance.
The detection of homozygous variants was not covered specifically in this validation, but it has been proved in principle by others (26). To detect homozygotes, one must mix samples in equal proportion with the wild-type control to allow the formation of heteroduplexes. Such mixing means that any heterozygous variant would be represented by only 25% of the sample. We have shown that the sensitivity of CSCE remains optimal only to approximately 22% representation of the mutant allele; sensitivity falls sharply below this value. It may therefore be prudent to use a lower proportion ofthe wild-type spike to keep the analysis of heterozygotes within the optimal range (e.g., 35% wild-type spike and 65% test sample).
LIMIT OF DETECTION
The findings that sensitivity is maintained down to 22% variant representation and that some variants are still detectable at 6% indicate that CSCE maybe useful for the analysis of mitochondrial DNA, tumor samples, and mosaic mutations, for which mutant representation is expected to be <50%.
CSCE is critically reliant on the discrimination of peak morphology. Therefore, robust monitoring of the resolution, ideally in all capillaries in every run, is critical for maintaining diagnostic standards. This goal can be accomplished most simply by reserving one of the color channels (typically LIZ) for running a resolution control. This control should consist of an amplicon containing a variant (preferably with a low resolution score) that produces a peak of known shape. This control can be monitored for appreciable changes in shape between capillaries and from run to run. Capillary analyses with results that fall outside the predefined limits should be considered as having failed.
[FIGURE 5 OMITTED]
RESOLUTION SCORES AND SPECIFIC EXAMPLES
Fig. 5A shows the distributions of mean resolution scores for manual analyses across the 3 laboratories, and Fig. 5B shows the distributions of the corresponding automated analysis scores. The majority (81%) of the insertion and deletion variants showed evidence of all 4 possible duplex species. Cases in which <4 peaks were detectable can be explained by lowered resolution due to proximity to the end of a fragment, chance equality of electrophoretic mobility of duplex species, or simply that peaks migrated beyond the electrophoresis data-collection window. Cases with >4 detectable species cannot be explained easily, because only 4 species are possible (2 homoduplexes and 2 heteroduplexes). It is notable that 22 of the 402 different variants in this study showed evidence of more than the predicted 4 duplex species (maximum, 8 species). This phenomenon occurred more frequently with insertion and deletion variants (9%) and compound variants (25%) than with single-base substitutions (2%). This finding suggests that for some variants, particularly when >1 base is affected, one or more of the duplexes formed are unstable and capable of assuming >1 conformation, thereby yielding >1 peak.
Aside from these observations, we found no specific correlations between the nature of the variation and detection efficacy; however, it is worth noting that point variations producing very subtle shifts, which are the most likely candidates for false-negative calls, are very rare. Only 5 variants (1.2%) showed no detectable shouldering; these variants were detectable only by a change in peak shape. Of these 5 variants, only 2 (<0.5%) appreciably challenged the automated analysis (i.e., they were not detected by both SRMS and MAXDIFF in all laboratories: BC1_16:c.4676-34G>Aand BC1_20:c.5194-312G>A).
Various special cases were investigated during the workup for the validation (Fig. 3). To evaluate the suggestion that the presence of a polymorphism in the same fragment as a pathogenic mutation could confound its detection, we included 16 different compound variants in the study. Seven of these compound variants consisted of a point variant juxtaposed with a small insertion or deletion; the remaining 9 variants were pairs of point variants. In all cases, the compound variant showed numbers of peaks equivalent to or greater than the numbers produced by the individual variants, and the profiles could easily be distinguished from the individual variants where they were available for analysis. Although the numbers are not large enough to draw any statistically significant conclusions, the presence of a polymorphism appears, if anything, to render true variants more detectable.
In cases in which the same variant was present in > 1 fragment (n = 34), the paired analyses yielded substantially different profiles. Similarly, in a single case in which 2 different base changes were located at the same nucleotide position, the 2 variants were clearly distinguishable (Fig. 3).
One variant, BRCA1 :c.5369_5385del17 in exon 22, was given a resolution score of 2 by manual analysis in laboratory 3, but it actually had a 4-peak profile. Owing to the separation of these peaks, only 2 peaks fell within the analysis window used. A similar error may have occurred in the analysis of BRCA1 :c.2646_2648delTGC in exon 11, which was one of the variants missed in laboratory 3. It is therefore important that manual analysis be performed in 2 stages: an overall view to detect large shifts, followed by detailed examination of the primary peak in an analysis window 200-500 scans wide to detect any subtle shifts.
Because of the width of the analysis window on the 3730 capillary arrays, it is necessary to include sucrose in the polymer to modify its refractive index so that all capillaries yield a suitable signal. This measure is not necessary on the 3130x/, although inclusion of sucrose did not detrimentally affect the analysis. The only important difference in the performance of the 2 instruments was optimizing the dilution and loading conditions so that all analyses fell within a suitable analysis window (500-6000 RFU for the 3130x/ and 1000-30 000 RFU for the 3730). This optimization requires substantially more effort for the 3130x/.
DEALING WITH NONPATHOGENIC POLYMORPHISMS
A number of variants in this study with low resolution scores gave very similar profiles that could not be reliably distinguished (Fig. 3). Therefore, because CSCE is not capable of unequivocally characterizing the variants detected, for diagnostic purposes all positive results, including those representing nonpathogenic polymorphisms, need to be confirmed by sequencing. This requirement can have a substantial effect on the cost and processing-time advantages gained by the use of CSCE. Alternative methods involving high-throughput or multiplex genotyping can be used to exclude known polymorphisms. For example, the Salisbury high-throughput screening facility uses a bespoke SNPlex assay (Applied Biosystems) to screen for 22 known polymorphisms in the breast cancer genes and has achieved an overall sequencing rate consistently <10% of the total fragments analyzed (unpublished data).
MATERIALS AND METHODOLOGY COMPARISON
For high-throughput diagnostic application, both CSCE and high-resolution melting analysis would be suitable candidates for a presequencing screen. The 2 methods are likely to have similar absolute sensitivities, although there may be differences with respect to the type of variants most likely to be called as false negatives (Supplementary Data I in the online Data Supplement). With respect to other aspects, there is little difference between the 2 methodologies. A key consideration in terms of methodology choice is likely to be the availability of instrumentation and the level of existing experience with the technologies.
In this study, we have defined optimal running conditions for applying CSCE to fragments with 20%-60% G + C content, which cover approximately 77% of all exons in the genome. We have derived a set of analysis guidelines for both manual and automated analyses. Although manual analysis can provide an effective mutation-scanning solution when the analysis guidelines are carefully applied, automated analysis is much faster and reduces the chances of operator error. With the use of multiplexing and automated analysis in the BioNumerics software, CSCE is substantially faster and cheaper than sequencing. This methodology is robust and relatively simple to implement for diagnostic applications. Furthermore, under the conditions defined and within the limitations described, CSCE with automated analysis provides a mutation-scanning technique for heterozygous nucleotide substitutions and small insertions and deletions that is, for practical purposes, equivalent to capillary-based Sanger sequencing in terms of sensitivity.
At the time of writing, Applied Biosystems was preparing a technical-application note regarding CSCE. This note will include a "troubleshooting" section. For any queries regarding CSCE or requests for specialist firmware and analysis modules mentioned in this report, please contact J. Theelen at Joop.Theelen@lifetech.com.
Author Contributions: All authors confirmed they have contributed to the intellectual content of this paper and have met the following 3 requirements: (a) significant contributions to the conception and design, acquisition of data, or analysis and interpretation of data; (b) drafting or revising the article for intellectual content; and (c) final approval of the published article.
Authors' Disclosures of Potential Conflicts of Interest: Upon manuscript submission, all authors completed the Disclosures of Potential Conflict of Interest form. Potential conflicts of interest:
Employment or Leadership: B. Pot, Applied Maths.
Consultant or Advisory Role: None declared.
Stock Ownership: None declared.
Honoraria: None declared.
Research Funding: The project was jointly funded and supported by the UK Department of Health (via NGRL Wessex), EuroGentest, Applied Biosystems (Europe), and Applied Maths (Gent, Belgium). Expert Testimony: None declared.
Role of Sponsor: The funding organizations played no role in the design of study, choice of enrolled patients, review and interpretation of data, or preparation or approval of manuscript.
Acknowledgments: We thank N. Arts and H. Ouchene for technical assistance.
(1.) Provaznikova D, Kumstyrova T, Kotlin R, Salaj P, Matoska V, Hrachovinova I, Rittich S. Highresolution melting analysis for detection of MYH9 mutations. Platelets 2008;19:471-5.
(2.) National Genetics Reference Laboratory - Wessex. Model for calculating the relative value of a pre-screen strategy with respect to sequencing. http://www.ngrl.org.uk/Wessex/downloads_reports. htm (Accessed March 2010).
(3.) Miki Y, Swensen J, Shattuck-Eidens D, Futreal PA, Harshman K, Tavtigian S, et al. A strong candidate for the breast and ovarian cancer susceptibility gene BRCA1. Science 1994;266:66-71.
(4.) Tavtigian SV, Simard J, Rommens J, Couch F, Shattuck-Eidens D, Neuhausen S, et al. The complete BRCA2 gene and mutations in chromosome 13q-linked kindreds. Nat Genet 1996;12:333-7.
(5.) National Human Genome Research Institute [NHGRI]. An open access on-line breast cancer mutation data base: an international collaborative effort hosted by NHGRI. http://research. nhgri.nih.gov/bic/ (Accessed March 2010).
(6.) Glavac D, Dean M. Applications of heteroduplex analysis for mutation detection in disease genes. Hum Mutat 1995;6:281-7.
(7.) van der Hout AH, van den Ouweland AM, van der Luijt RB, Gille HJ, Bodmer D, Bmggenwirth H, et al. A DGGE system for comprehensive mutation screening of BRCA1 and BRCA2: application in a Dutch cancer clinic setting. Hum Mutat 2006;27: 654-66.
(8.) Friedman LS, Szabo CI, Ostermeyer EA, Dowd P, Butler L, Park T, et al. Novel inherited mutations and variable expressivity of BRCA1 alleles, including the founder mutation 185delAG in Ashkenazi Jewish families. Am J Hum Genet 1995;57:1284-97.
(9.) Kozlowski P, Krzyzosiak WJ. Combined SSCP/duplex analysis by capillary electrophoresis for more efficient mutation detection. Nucleic Acids Res 2001;15:29:E71.
(10.) Gayther SA, Harrington P, Russell P, Kharkevich G, Garkavtseva RF, Ponder BA. Rapid detection of regionally clustered germ-line BRCA1 mutations by multiplex heteroduplex analysis. UKCCCR Familial Ovarian Cancer Study Group. Am J Hum Genet 1996;58:451-6.
(11.) Ganguly T, Dhulipala R, Godmilow L, Ganguly A. High throughput fluorescence-based conformation-sensitive gel electrophoresis (F-CSGE) identifies six unique BRCA2 mutations and an overall low incidence of BRCA2 mutations in high-risk BRCA1negative breast cancer families. Hum Genet 1998; 102:549 -56.
(12.) Ganguly A. An update on conformation sensitive gel electrophoresis. Hum Mutat 2002;19:334-42.
(13.) Marsh DJ, Theodosopoulos G, Howell V, Richardson AL, Benn DE, Proos AL, et al. Rapid mutation scanning of genes associated with familial cancer syndromes using denaturing high-performance liquid chromatography. Neoplasia 2001;3:236-44.
(14.) Lu AL, Hsu IC. Detection of single DNA base mutations with mismatch repair enzymes. Genomics 1992;14:249-55.
(15.) Gerhardus A, Schleberger H, Schlegelberger B, Gadzicki D. Diagnostic accuracy of methods for the detection of BRCA1 and BRCA2 mutations: a systematic review. Eur J Hum Genet 2007;15: 619-27.
(16.) De Leeneer K, Coene I, Poppe B, De Paepe A, Claes K. Rapid and sensitive detection of BRCA1/2 mutations in a diagnostic setting: comparison of two high-resolution melting platforms. Clin Chem 2008;54:982-9.
(17.) van der Stoep N, van Paridon CDN, Janssens T, Krenkova P, Stambergova A, Macek M, et al. Diagnostic guidelines for high-resolution melting curve (HRM) analysis: an interlaboratory validation of BRCA1 mutation scanning using the 96well LightScanner. Hum Mutat 2009;30:899-909.
(18.) Vossen RH, Aten E, Roos A, den Dunnen JT. High-resolution melting analysis (HRMA): more than just sequence variant screening. Hum Mutat 2009;30:860-6.
(19.) Esteban-Cardenosa E, Duran M, Infante M, Velasco E, Miner C. High-throughput mutation detection method to scan BRCA1 and BRCA2 based on heteroduplex analysis by capillary array electrophoresis. Clin Chem 2004;50:313-20.
(20.) Velasco E, Infante M, Duran M, Esteban-Cardenosa E, Lastra E, Garcia-Giron C, Miner C. Rapid mutation detection in complex genes by heteroduplex analysis with capillary array electro phoresis. Electrophoresis 2005;26:2539-52.
(21.) Davies H, Dicks E, Stephens P, Cox C, Teague J, Greenman C, et al. High throughput DNA sequence variant detection by conformation sensitive capillary electrophoresis and automated peak comparison. Genomics 2006;87:427-32.
(22.) Ashton EJ, Yau SC, Deans ZC, Abbs SJ. Simultaneous mutation scanning for gross deletions, duplications and point mutations in the DMD gene. Eur J Hum Genet 2008;16:53-61.
(23.) Applied Maths. CSCE-based hetero-duplex analysis (HDA) using the BioNumerics HDA Plugin. http://www.applied-maths.com/bionumerics/plug ins/hda.htm (Accessed March 2010).
(24.) National Genetics Reference Laboratory (Wessex). Plasmid based generic mutation detection reference reagents; production and performance indicator field trial. http://www.ngrl.org.uk/Wessex/ downloads/pdf/Generic.pdf (Accessed March 2010).
(25.) National Genetics Reference Laboratory (Wessex). Standardised primer optimisation and design specification. http://www.ngrl.org.uk/Wessex/ downloads/pdf/NGRLW_SPODS_2.2.pdf (Accessed March 2010).
(26.) van den Akker P, Hettema W, Meijer R, Jonkman M, Hofstra R, Scheffer H. Design and validation of a conformation-sensitive capillary electrophoresis system for mutation identification of the COL7A1 gene with automated peak comparison. Genet Test Mol Biomarkers 2009;13:589-97.
Christopher J. Mattocks,  * Gemma Watkins,  Daniel Ward,  Tom Janssens,  Ermanno A.J. Bosgoed,  Kim van der Donk,  Marjolijn J. Ligtenberg,  Bruno Pot,  Joop Theelen,  Nicholas C.P. Cross,  Hans Scheffer,  and Gert Matthijs 
 National Genetics Reference Laboratory (Wessex), Salisbury, UK;  Wessex Regional Genetics Laboratory, Salisbury, UK;  Center for Human Genetics, Leuven, Belgium;  Department of Human Genetics, Radboud University Nijmegen Medical Centre, Nijmegen, the Netherlands;  Applied Maths, Gent, Belgium;  Applied Biosystems, Nieuwerkerk aan den lJssel, the Netherlands.
 Human genes: BRCA1, breast cancer 1, early onset; BRCA2, breast cancer 2, early onset.
 Nonstandard abbreviations: CSCE, conformation-sensitive capillary electrophoresis; RFU, relative fluorescence units; SRMS, single-sided root mean square; MAXDIFF, largest vertical difference between 2 compared curves; H0, null hypothesis.
* Address correspondence to this author at: National Genetics Reference Laboratory (Wessex), Salisbury District Hospital, Salisbury SP2 8BJ, UK. Fax +441722-338095; e-mail firstname.lastname@example.org.
Received August 11, 2009; accepted January 22, 2010.
Previously published online at DOI: 10.1373/clinchem.2009.135426
Table 1. Summary results for methodology validation. Lab 1 Lab 2 Lab 3 All Automated analysis, n True positive 385 371 390 1146 False negative 0 0 1 1 True negative 351 292 284 927 False positive 7 33 44 84 Analysis fails 24 71 48 143 Controls (excluded) (a) 150 150 150 450 Replicates (excluded) (b) 43 43 43 129 Total analyses 960 960 960 2880 Sensitivity, % 100 100 99.7 99.9 Lower 95% confidence 99.2 99.2 98.8 99.6 limit (c) Upper 95% confidence 100 100 100 100 limit (c) Specificity, % 98 89.8 86.6 91.7 Lower 95% confidence 96.4 86.7 83.1 90.1 limit (c) Upper 95% confidence 99.1 92.3 89.6 93.1 limit (c) Manual analysis, n True positive 394 373 366 1133 False negative 0 1 11 12 True negative 346 303 310 958 False positive 16 36 5 57 Analysis fails 11 54 75 140 Controls (excluded) (a) 150 150 150 450 Replicates (excluded)b 43 43 43 129 Total analyses 960 960 960 2880 Sensitivity, % 100 99.7 97.1 99 Lower 95% confidence 99.2 98.7 95.2 98.3 limit (c) Upper 95% confidence 100 100 98.4 99.4 limit (c) Specificity, % 95.6 89.4 98.4 94.4 Lower 95% confidence 93.4 86.2 96.7 93.1 limit (c) Upper 95% confidence 97.2 92 99.4 95.5 limit (c) (a) Defined controls for each fragment are excluded from the calculation of specificity. (b) Some variants were present in >1 analysis. Such replicates were excluded from the sensitivity calculations (i.e., only 1 copy of each variant analysis was considered). Replicates for exclusion were selected randomly and were the same for all analyses. (c) Confidence limits were calculated by the exact method for small numbers of failures (see http://www.itl.nist.gov/div898/ handbook/prc/section2/prc241.htm).
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||Molecular Diagnostics and Genetics|
|Author:||Mattocks, Christopher J.; Watkins, Gemma; Ward, Daniel; Janssens, Tom; Bosgoed, Ermanno A.J.; van de|
|Date:||Apr 1, 2010|
|Previous Article:||Correlation between methylphenidate and ritalinic acid concentrations in oral fluid and plasma.|
|Next Article:||MicroRNA-21 is induced early in pancreatic ductal adenocarcinoma precursor lesions.|