Analysis of molecular inversion probe performance for allele copy number determination
- Yuker Wang†1,
- Martin Moorhead†1,
- George Karlin-Neumann1,
- Nicholas J Wang2,
- James Ireland1,
- Steven Lin1,
- Chunnuan Chen1,
- Laura M Heiser2,
- Koei Chin3,
- Laura Esserman3,
- Joe W Gray2,
- Paul T Spellman2Email author and
- Malek Faham1Email author
© Wang et al.; licensee BioMed Central Ltd. 2007
Received: 6 July 2007
Accepted: 20 November 2007
Published: 20 November 2007
We have developed a new protocol for using molecular inversion probes to accurately and specifically measure allele copy number. The new protocol provides for significant improvements, including the reduction of input DNA (from 2 μg) by more than 25-fold (to 75 ng total genomic DNA), higher overall precision resulting in one order of magnitude lower false positive rate, and greater dynamic range with accurate absolute copy number up to 60 copies.
Chromosomal copy number analysis has been important in the study of tumor samples for decades. Changes in copy number have already been demonstrated to predict patients' response and/or prognosis , which gives hope that this can be applied in large scale to significantly affect clinical care in the future. In order to fulfill this promise, technologies that are able to assess copy number on the whole genome scale in a large number of samples are required. Since the development of comparative genomic hybridization (CGH) , many technologies have been developed to address this need. These include bacterial artificial chromosome (BAC) CGH and, more recently, CGH employing several types of oligonucleotides arrays [3–7]. Some of the newer CGH methodologies allow for allelic information to be obtained [4, 5, 7, 8]. The utility of measurement of allele copy number (ACN) includes the identification of loss of heterozygosity (LOH) events  and the allelic composition at amplified loci .
One of the techniques that have previously been described for the measurement of ACN is molecular inversion probes (MIPs) [10–12]. Briefly, MIP probes are circularizable oligonucleotides, where the two ends carry two sequences that are complementary to two sequences on the genome separated by one nucleotide (exactly where the variant to be genotyped is). After hybridization to the genomic DNA, the reaction is split into four tubes where a single nucleotide is added to each tube. Upon the addition of the nucleotide, the MIP probe is ligated closed (but this only occurs in the tube with the nucleotide that is complementary to the allele on the genome), turning the probe into a circle. This structure can be selected for by the use of exonucleases, allowing for minimal 'cross talk' between probes and making it possible to obtain high quality data from highly multiplexed assays (>50,000-plex). Ultimately, these products are amplified and hybridized onto an Affymetrix microarray to identify the present products.
The MIP assay differs from other highly multiplexed (tens of thousands to hundreds of thousands) genotyping techniques in that it utilizes enzymatic steps in solution to capture specific loci, which is then followed by an amplification step. Such a combination of enzymatic steps confers a high degree of specificity on the MIP assay. The high specificity and minimum 'cross talk' between loci or alleles results in precise measurements as well as large assay dynamic range. In addition, the amplification of the loci of interest only simplifies the task of detection and provides the ability to use lower amounts of input genomic DNA. The high precision, large dynamic range, and low DNA usage are demonstrated in this study. Finally, because MIP requires only 40 base-pairs of intact genomic DNA, its use in degraded samples, such as formaldehyde fixed paraffin embedded samples, may offer distinct advantages.
We have made significant advancements in this technology. As a result, the false positive rate has decreased by an order of magnitude and the dynamic range extended to achieve accurate absolute copy number measurements up to 60 copies, while reducing the input genomic DNA requirement by more than 25-fold.
We describe the performance of the MIP assay using several types of metrics that are broadly useful to all copy number assays: the ability to discriminate a copy number aberration from normal at the total as well as ACN level; and the ability to accurately quantify the level of copy number aberration at both the total and ACN levels.
MIP copy number assay modification
We have previously described the use of MIP for copy number analysis [11, 12]. We have now improved the performance of the technology through modifications of the MIP copy number protocol and through improved data analysis. The improved performance allows ACN data to be obtained using 75 ng of human genomic DNA.
The first implementation of the MIP ACN assay required 2 mg of genomic DNA. We discovered that only a fraction of the genomic templates hybridized to MIP probes that are then circularized and amplified. We hypothesized that increasing the number of MIP molecules and decreasing the hybridization volume should increase the number of MIP molecules bound to their genomic targets. We tested this hypothesis and verified that increasing the number of MIP molecules by a factor of four and decreasing the hybridization volume (from 27 ml to 6.7 ml) allowed us to substantially decrease genomic DNA input. After the hybridization, buffer is added to increase the volume to 27 ml, and the rest of the protocol is unmodified.
In the standard genotyping protocol, the genomic target is split into four reactions, where one of each of the four nucleotides is added. We recognized that we could decrease DNA input requirements by performing a smaller number of these reactions We reasoned that if we were to use only one set of single nucleotide polymorphisms (SNPs; for example only the most common C/T SNPs), we would decrease the DNA requirement by 50%. Similarly, adding two nucleotides into each of two reactions leads to the same result. We have implemented this variant protocol by adding G and C nucleotides into one tube, and adding A and T into another. In this scenario, about 85% of SNPs in the human genome (all but G/C and A/T SNPs) can be assessed. An advantage of decreasing the number of reactions is that it requires only two independent readouts rather than four (that is, four colors on one array or one color on four arrays). In the optimized procedure, 75 ng of genomic DNA are mixed with more than 50,000 probes in a small volume (6.7 ml). The hybridized probe:target genomic DNA are split into two reactions, where two nucleotides are added to each of the two tubes. The two reactions are processed separately and read on two independent arrays, which was found to yield better data than two colors on one array (data not shown).
One effect that requires correction in quantitative assays on arrays is the phenomenon of saturation. This is especially important for correct estimations of amplifications. We have implemented a Langmuir correction for the non-linear relationship between signal and copy number . Our algorithm was developed on a separate data set, and the data shown here is an independent set. Using this algorithm we have been able to measure copy number in a linear fashion at levels over 60 copies (see below).
Detection of aberrations
An important aspect of the copy number performance is the detection of aberrations where the copy number is distinct from 2. The degree of discrimination between copy number 2 and the aberrant copy can be understood through receiver operator characteristic (ROC) curves showing the trade off between false positive rate and sensitivity (1 - false negative rate) given data on regions with known copy number. The presence of cell lines carrying 1, 3, 4, or 5 X chromosomes provides a good resource for the study of the performance of the technology in this copy number range . For example, in the assessment of cell lines with one X chromosome (males) one can make a threshold at copy number 1.5 and any marker on the X chromosome with a copy number below 1.5 would be considered a true positive, and any autosomal marker with a copy number below 1.5 is considered a false positive. By plotting this trade off between true and false positives at many thresholds between copy numbers of 0 to 3, the full ROC curve is generated.
Systematic false positives
The above analysis assumes that all the autosomal markers are present at two copies per cell. There has been a wealth of evidence demonstrating copy number polymorphisms (CNPs) in the general population [14, 15]. Therefore, a fraction of what we considered as false positives may in fact be true positives. In addition, the presence of a secondary SNP (distinct from the one being interrogated) within the probe may emulate the presence of a deletion. Data generated on two CEPH pedigree populations, Yoruban and Utah, are informative in this regard because the polymorphisms on which the MIP panel is based are from European (equivalent to Utah) rather than African populations. The contribution of genetic variants (CNP or SNP) to the apparent false positive rate is suggested by our detection of approximately three-fold more apparent autosomal deletions in the Yoruban population compared to the Utah population (average number of markers per sample with measured copy number below 1.3 is 126 markers for the Utah population and 319 for the Yoruban population). We hypothesized that this imbalance between the number of apparent deletions in the two populations was likely due to secondary polymorphisms close to the SNP being assayed, which prevent proper binding of the MIP to its target. Further evidence to support this hypothesis was noted when we observed that the majority of these apparent deletions were reproducible when a sample is re-assayed.
To understand the nature of these apparent deletions, we randomly picked nine SNPs, which showed copy number measurements below 1.3 in replicate measurements from the Yoruba sample (sample NA18515). We PCR amplified approximately 400 base-pair fragments that included the SNP assayed by MIP and used dideoxy sequencing to show that eight of these nine loci that were successfully sequenced had a secondary SNP within the MIP probe homology sequence. The ninth SNP that showed copy number 1 was assayed by qPCR to measure copy number but was found to show a normal copy number of two (Supplementary Table 1 in Additional data file 1).
Trade off between resolution and performance
Copy number changes are expected to occur in discrete segments, allowing neighboring markers to be averaged together. This leads to enhanced performance as measured by the trade off between false positive rate and sensitivity (that is, the ROC curve moving to the upper left) at the expense of lower resolution.
As discussed above, one shortcoming of the ROC analysis is the presence of CNPs in the autosomes. Averaging two adjacent markers that lie within a CNP will erroneously consider these markers as false. Therefore, for the purpose of describing the performance of the technology, we averaged markers that are not adjacent to each other. This method would ameliorate the effect of miscalling two adjacent markers in a CNP as a false positive. This analysis is appropriate as long as there is a lack of correlation between marker performance and the position on the chromosome. If this assumption is true, then the operation reflects the performance of averaging two adjacent markers since the adjacent and the random markers are obtained from the same distribution. Clearly, averaging data from non-adjacent markers is valid only for the assessment of the technology performance and cannot generate any meaningful biological findings.
Sensitivity at 50% specificity
40 K (75% of data)
48 K (90% of data)
Accuracy of copy number estimation
Expected versus measured copy number
Expected copy number
Measured copy number
Relative standard deviation
Allele copy number in spiked samples
Identification of LOH without matched normal tissue
A major challenge in the study of ACN is the absence of matched normal tissue for many valuable clinical samples. In tumors that have lost one allele, it is not easy to discriminate LOH for individual alleles that are homozygous in the entire individual. We recognized that the high sensitivity and accuracy of the MIP ACN assay, coupled with the high likelihood of normal tumor contamination, could allow us to distinguish LOH from alleles that are homozygous. In theory, this should be best accomplished with tumor showing substantial (approaching 50%) normal contamination.
Genotyping disagreements between replicated samples
We describe in this manuscript significant improvements we have made to the MIP-based measurements of ACN. By increasing the proportion of genomic targets that are hybridized to the MIP probes, we have improved the performance while requiring a smaller amount of DNA. Additionally, for copy number measurements there are substantial advantages in uniformity and robustness when utilizing one-color readouts, especially at high levels of multiplexing. The use of a control sample that is co-hybridized with the test sample in an analogous fashion as used by BAC arrays leads to inferior results compared with the one color readout (data not shown). Presumably, this is because the different dyes have different characteristics of brightness and saturation. We conclude that the effect of the lack of uniformity among the dyes is probably larger in our system than chip-to-chip variation that the control sample co-hybridization is supposed to ameliorate. The improvements achieved from the new protocol as evaluated by ROC curve analysis resulted in a decline in the false positive rate by an order of magnitude, while reducing the input genomic DNA by more than 25-fold. In addition, the dynamic range has been extended with accurate estimation achieved for up to 60 copies.
We evaluated the performance of MIP for ACN measurements using a set of metrics that are broadly useful for all copy number assays. We demonstrate the ability of MIP to detect a single copy deletion or duplication at an allele and total copy number levels using ROC curve analysis. We believe ROC curve analysis provides a rigorous statistical framework for comparing different technologies or different protocols/algorithms of the same fundamental technology. In addition to genuinely improving the technology performance in the ROC curves by the use of better protocol and algorithms, one may apparently improve them by smoothing (Figure 5), or filtering the worst markers (Figure 5) or the worst samples (Figure 3).
We have shown in the single MIP marker analysis that many of the apparent false positives in the discrimination between 1 and 2 copies are due to the presence of SNPs in the genomic sequence that are complementary to the MIP probes. This effect will be strongest in populations that are the most diverse. It should be possible to ameliorate this effect by using matched normal and tumor pairs. The presence of SNPs may explain why the discrimination between 1 and 2 is not better than that between 2 and 3, as secondary SNPs that interfere with MIP binding emulate a copy number deletion.
We also show the MIP assay precision of measurements of copy number at allele and total copy number levels. Precision at the total copy number level requires low background of the assay and lack of saturation. In addition, allele level precision requires a low level of allele cross talk even when one allele is present in huge excess relative to the other.
These observations led us to suspect that it should be possible to genotype mixed DNA populations, such as occurs in tumor samples contaminated with normal tissue. As normal contamination increases, some estimate of the amount of normal contamination is valuable, which we believe can be quite accurately estimated using the calculated copy numbers for regions of LOH and deletion.
One promise of ACN data over the traditional total copy number data is the potential that it may facilitate the identification of the critical genes in regions of aberrations. Even though large aberrations can be readily identified by total copy number CGH, the identification of the critical gene(s) in these aberrations is often not straightforward. This is in contrast to sequencing data where identification of mutations has been quite laborious, but once achieved the critical gene is usually easily identified. Identification of an allele that is preferentially deleted or amplified in a set of samples implicates the specific allele (or one in linkage disequilibrium with it) as critical in the pathogenesis of the aberrations.
Materials and methods
Samples and MIP assay
The normal samples as well as the samples carrying 3 (NA04626), 4 (NA01416), and 5(NA06061) copies of the X chromosome were obtained from Coriell Cell Repository (Camden, NJ, USA). The normal HapMap samples that were used were also obtained from Coriell Cell Repository. The samples that were used were: NA19240, NA19239, NA06991, NA06985, NA19238, NA19222, NA19202, NA19201, NA19200, NA19132, NA19131, NA18956, NA18951, NA18949, NA18947, NA18945, NA18912, NA18854, NA19130, NA19128, NA19127, NA19099, NA19094, NA18991, NA18987, NA18981, NA18605, NA18603, NA18582, NA18573, NA18558, NA18550, NA18547, NA18542, NA18537, NA18515, NA18508, NA12892, NA12813, NA12717, NA12156, NA12155, NA12004, NA11881, NA11840, NA11832, NA11830, NA10831, NA07345, NA07056, NA07029, NA07019, NA07000, and NA06993. The MCF7 cell line was obtained from the American Tissue Cell Culture (ATCC, Manassas, VA, USA).
The MIP assay was performed as described previously, but with important modifications . Specifically, the current protocol is a modification of the targeted genotyping protocol commercialized by Affymetrix (additional information about MIP technology can be found at the Affymetrix website ). Briefly, test DNA samples were diluted to 16 ng/ml. All DNA quantification was done using PicoGreen dsDNA Assay Kit (Molecular Probes/Invitrogen, Carlsbad, CA, USA, P7589). We used 96- or 384-well plates whenever possible to reduce variation. For day1 overnight annealing, 4.7 μl of DNA samples (75 ng total), 0.75 μl of Buffer A, 1.1 μl of the 53 K probe pool (200 amol/μl/probe) and 0.045 μl of Enzyme A were mixed well in a 384-well plate on ice. The reaction was incubated at 20°C for 4 minutes, 95°C 5 minutes, then 58°C overnight. On day 2, 13 μl of Buffer A was added to each well with 1.25 μl of Gapfill Enzyme mix. Then, 9 μl of this was put in each of two wells in a 96-well plate. MIP probes were circularized with 4 μl of dinucleotide (dATP with dTTP, dCTP with dGTP) and mixed at 58°C for 10 minutes. The uncircularized probes and genomic DNA were eliminated by addition of 4 μl of Exonuclease Mix and incubation at 37°C for 15 minutes, followed by heat-killing of enzymes. The circularized probes were linearized by the addition of Cleavage Enzyme Mix at 37°C for 15 minutes, then subjected to universal primer amplification for 18 cycles at 95°C for 20 s, 64°C for 40 s and 72°C for 10 s. For the labeling reaction, the product was further amplified with the label primers for 10 cycles, and then subjected to cleavage by Digest Enzyme Mix at 37°C for 2 h. To hybridize, the cleaved MIP products were mixed with hybridization cocktail, denatured and hybridized to 70 K Universal Taq arrays at 39°C for 16 h (two arrays per sample). The overnight hybridized arrays were washed on GeneChip® Fluidics Station FS450 and stained by SAPE at 5 ng/ml (Invitrogen).
Copy number estimation was obtained from the hybridization signals as described previously, but with the following modifications . Given that in this work no multi-color readout was present (but rather single color readout on two arrays), no spectral overlap was present and, therefore, the color-seperation step was omitted. In addition, instead of the linear calibration of the allele signals, Langmuir correction was done .
Generation of spike-in samples
A panel of 80 PCR products representing genomic regions containing MIPs on chromosome 2 were PCR amplified from CEPH1341.14 (NA06985) using an ABI 9700 thermocycler (initial denaturation of 95°C for 5 minutes, 95°C for 30 s, 58°C for 30 s, 72°C for 60 s for 30 cycles; final extension at 72°C for 7 minutes). The products were purified using a MinElute 96 UF PCR Purification plate (Qiagen Valencia, CA, USA) and resuspended in TE. The purified products were quantified on a fluorometer using the Quant-It™ dsDNA Assay kit (Invitrogen). Purified PCR products were then pooled into ten tubes, each containing eight different products (Supplementary Table 2 in Additional data file 1). Each pooled tube of probes was then serially diluted two-fold into a series of spike-in tubes containing 150 ng of genomic DNA from CEPH1341.02 (NA06991) (Supplementary Table 2 in Additional data file 1). The genomic DNA samples were chosen so that the spike-in PCR products from CEPH1341.14 represented a single allele, while the genomic DNA from CEPH 1341.02 was heterozygous, allowing for discrimination of allele specific amplification.
Sequence analysis of aberrant MIPs
PCR products were amplified using primers designed to span sequences containing MIPs that did not hybridize as expected (Supplementary Table 1 in Additional data file 1). Amplification was carried out in a 50 μl reaction (initial denaturation of 95°C for 5 minutes, 95°C for 30 s, 58°C for 30 s, 72°C for 60 s for 30 cycles; final extension at 72°C for 7 minutes) and products were purified using a MinElute 96 UF PCR Purification plate (Qiagen) and resuspended in TE. The purified products were sequenced using an Applied Biosystems (Foster City, CA, USA) 96 capillary 3730 × l DNA Analyzer and the forward and reverse primers used during amplification.
Identification of LOH without matched normal tissue
Genotyping metrics from the traditional MIP method were applied to each observation and estimated genotypes (AA, AB, or BB) were determined for each MIP in each of five replicated tumor samples. Data are provided as Additional data file 1. Regions of the genome that show clear evidence for decreases in copy number are easily observed with the decrease in copy number equivalent to 1.5 total copies (1 copy of 1 allele and 0.5 copies of the other, or for homozygous alleles 1.5 total copies). No regions of the genome in any of the five samples analyzed appear to have ~1 copy of the higher allele and ~0 copies of the lower allele.
Additional data files
The following additional data are available with the online version of this paper. Additional data file 1 lists replicated ACN data for five breast cancers. Data are filtered to use only high quality MIPs (90% or greater call rate, less than 12% copy number variation).
allele copy number
bacterial artificial chromosome
comparative genome hybridization
copy number polymorphism
loss of heterozygosity
molecular inversion probe
receiver operator characteristic
single nucleotide polymorphism.
This work was supported by Affymetrix and NIH award U54CA112970 to JWG. Disclaimer: this document was prepared as an account of work sponsored by the United States Government. While this document is believed to contain correct information, neither the United States Government nor any agency thereof, nor The Regents of the University of California, nor any of their employees, makes any warranty, express or implied, or assumes any legal responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed, or represents that its use would not infringe privately owned rights. Reference herein to any specific commercial product, process, or service by its trade name, trademark, manufacturer, or otherwise, does not necessarily constitute or imply its endorsement, recommendation, or favoring by the United States Government or any agency thereof, or The Regents of the University of California. The views and opinions of authors expressed herein do not necessarily state or reflect those of the United States Government or any agency thereof, or The Regents of the University of California.
- Slamon DJ, Leyland-Jones B, Shak S, Fuchs H, Paton V, Bajamonde A, Fleming T, Eiermann W, Wolter J, Pegram M, et al: Use of chemotherapy plus a monoclonal antibody against HER2 for metastatic breast cancer that overexpresses HER2. N Engl J Med. 2001, 344: 783-792. 10.1056/NEJM200103153441101.PubMedView ArticleGoogle Scholar
- Kallioniemi OP, Kallioniemi A, Sudar D, Rutovitz D, Gray JW, Waldman F, Pinkel D: Comparative genomic hybridization: a rapid new method for detecting and mapping DNA amplification in tumors. Sem Cancer Biol. 1993, 4: 41-46.Google Scholar
- Lucito R, Healy J, Alexander J, Reiner A, Esposito D, Chi M, Rodgers L, Brady A, Sebat J, Troge J, et al: Representational oligonucleotide microarray analysis: a high-resolution method to detect genome copy number variation. Genome Res. 2003, 13: 2291-2305. 10.1101/gr.1349003.PubMedPubMed CentralView ArticleGoogle Scholar
- Bignell GR, Huang J, Greshock J, Watt S, Butler A, West S, Grigorova M, Jones KW, Wei W, Stratton MR, et al: High-resolution analysis of DNA copy number using oligonucleotide microarrays. Genome Res. 2004, 14: 287-295. 10.1101/gr.2012304.PubMedPubMed CentralView ArticleGoogle Scholar
- Zhao X, Li C, Paez JG, Chin K, Janne PA, Chen TH, Girard L, Minna J, Christiani D, Leo C, et al: An integrated view of copy number and allelic alterations in the cancer genome using single nucleotide polymorphism arrays. Cancer Res. 2004, 64: 3060-3071. 10.1158/0008-5472.CAN-03-3308.PubMedView ArticleGoogle Scholar
- Barrett MT, Scheffer A, Ben-Dor A, Sampas N, Lipson D, Kincaid R, Tsang P, Curry B, Baird K, Meltzer PS, et al: Comparative genomic hybridization using oligonucleotide microarrays and total genomic DNA. Proc Natl Acad Sci USA. 2004, 101: 17765-17770. 10.1073/pnas.0407979101.PubMedPubMed CentralView ArticleGoogle Scholar
- Ishikawa S, Komura D, Tsuji S, Nishimura K, Yamamoto S, Panda B, Huang J, Fukayama M, Jones KW, Aburatani H: Allelic dosage analysis with genotyping microarrays. Biochem Biophys Res Comm. 2005, 333: 1309-1314. 10.1016/j.bbrc.2005.06.040.PubMedView ArticleGoogle Scholar
- Peiffer DA, Le JM, Steemers FJ, Chang W, Jenniges T, Garcia F, Haden K, Li J, Shaw CA, Belmont J, et al: High-resolution genomic profiling of chromosomal aberrations using Infinium whole-genome genotyping. Genome Res. 2006, 16: 1136-1148. 10.1101/gr.5402306.PubMedPubMed CentralView ArticleGoogle Scholar
- Ewart-Toland A, Briassouli P, de Koning JP, Mao JH, Yuan J, Chan F, MacCarthy-Morrogh L, Ponder BA, Nagase H, Burn J, et al: Identification of Stk6/STK15 as a candidate low-penetrance tumor-susceptibility gene in mouse and human. Nat Genet. 2003, 34: 403-412. 10.1038/ng1220.PubMedView ArticleGoogle Scholar
- Hardenbol P, Yu F, Belmont J, Mackenzie J, Bruckner C, Brundage T, Boudreau A, Chow S, Eberle J, Erbilgin A, et al: Highly multiplexed molecular inversion probe genotyping: over 10,000 targeted SNPs genotyped in a single tube assay. Genome Res. 2005, 15: 269-275. 10.1101/gr.3185605.PubMedPubMed CentralView ArticleGoogle Scholar
- Ji H, Kumm J, Zhang M, Farnam K, Salari K, Faham M, Ford JM, Davis RW: Molecular inversion probe analysis of gene copy alterations reveals distinct categories of colorectal carcinoma. Cancer Res. 2006, 66: 7910-7919. 10.1158/0008-5472.CAN-06-0595.PubMedPubMed CentralView ArticleGoogle Scholar
- Wang Y, Moorhead M, Karlin-Neumann G, Falkowski M, Chen C, Siddiqui F, Davis RW, Willis TD, Faham M: Allele quantification using molecular inversion probes (MIP). Nucleic Acids Res. 2005, 33: e183-10.1093/nar/gni177.PubMedPubMed CentralView ArticleGoogle Scholar
- Burden CJ, Pittelkow YE, Wilson SR: Statistical analysis of adsorption models for oligonucleotide microarrays. Stat App Genet Mol Biol. 2004, 3: Article35-Google Scholar
- Sebat J, Lakshmi B, Troge J, Alexander J, Young J, Lundin P, Maner S, Massa H, Walker M, Chi M, et al: Large-scale copy number polymorphism in the human genome. Science. 2004, 305: 525-528. 10.1126/science.1098918.PubMedView ArticleGoogle Scholar
- Iafrate AJ, Feuk L, Rivera MN, Listewnik ML, Donahoe PK, Qi Y, Scherer SW, Lee C: Detection of large-scale variation in the human genome. Nat Genet. 2004, 36: 949-951. 10.1038/ng1416.PubMedView ArticleGoogle Scholar
- Affymetrix: Targeted Genotyping. [http://www.affymetrix.com/products/application/targeted_genotyping.affx]
This article is published under license to BioMed Central Ltd. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.