Ambry Genetics, a prominent DNA-testing firm located in Orange County, CA, unveiled a new databank on Tuesday containing the aggregated genetic information of 10,000 patients with hereditary breast and ovarian cancer. The free initiative named AmbryShare is open to the public, but the information is anonymized. It was engineered to help support President Obama’s Precision Medicine program.
News Article | March 8, 2016
Ambry Genetics is expected to announce on Tuesday that it will put information from 10,000 customers into a publicly available database.
No statistical methods were used to predetermine sample size. The experiments were not randomized. The investigators were not blinded to allocation during experiments and outcome assessment. NIH/3T3 tet-on 3G cells (Clontech, 631197) were cultured in DMEM (Invitrogen, 10566-016) supplemented with 10% FBS (Sigma, F4135-500ML) and 100 U ml−1 penicillin–streptomycin (Invitrogen, 15140-122). Mouse ESCs were cultured as described28. Single-cell suspension after trypsinization was used for 4′,6-diamidino-2-phenylindole (DAPI) staining immediately before sorting by flow cytometry. Single live cells were sorted and deposited directly into each tube of a PCR strip-tube, which contained 30 μl cell lysis buffer (10 mM Tris-HCl, pH 7.5, 10 mM NaCl, 3 mM MgCl , 0.1% Triton X-100). To prevent loss of the extremely small amount of DNase I hypersensitive DNA (<0.1 pg) released by DNase I digestion of single cells, we added a large amount of circular plasmid DNA (30 ng; about 3 × 105 times the amount of the DHS DNA in a single cell) as carrier DNA in the subsequent steps of library preparation. The circular DNA was not compatible with the adaptor ligation and thus could minimize the non-specific amplification by the subsequent PCR. The PCR conditions were optimized to amplify the small fragments (<200 base pairs (bp)) derived from DNase I hypersensitive sites without previous fractionation of these fragments. For DNase I digestion, 0.2 to 1 unit of DNase I (Roche, 04716728001) was added to the cells and incubated at 37 °C for 5 min. The reaction was stopped by adding 80 μl of stop buffer (10 mM Tris-HCl, pH 7.5, 10 mM NaCl, 0.15% SDS, 10 mM EDTA) containing 1 μl of 20 mg ml−1 proteinase K and 5 μl of 6 ng μl−1 circular carrier DNA. The mixture was incubated at 55 °C for 1 h and DNA purified by phenol–chloroform extraction, followed by precipitation with ethanol in the presence of 20 μg glycogen. The library was prepared using Illumina kits as described29. The libraries were amplified using a two-step method to preferentially amplify the small DNA fragments derived from the DNA hypersensitive sites and to reduce non-specific amplification of the carrier DNA. The first amplification was done with index primers with the PCR condition 98 °C for 10 s, 67 °C for 30 s, 72 °C for 30 s for six cycles. After isolation of the desired fragments (160–300 bp) using 2% E-gel (Invitrogen), the second amplification was done with the P5 and P7 primers with the condition 98 °C for 10 s, 68 °C for 30 s, 72 °C for 30 s for 22 cycles. The fragments between 160 and 300 bp were isolated on E-gel and sequenced on Illumina HiSeq 2500. The anonymized tumour samples from Ambry Genetics, approved by institutional review board and with informed consent, were used in this study. Three cases of thyroid cancer were diagnosed as FTC and one case was diagnosed as papillary thyroid carcinoma. Cells were manually scraped off from the highlighted area of a paraffin slide using a razor blade and resuspended in 150 μl of de-paraffinization solution (Qiagen, 1064343) and incubated at 56 °C for 3 min. After cooling to room temperature (about 25 °C), 150 μl of lysis buffer (10 mM Tris-HCl, pH 7.5, 10 mM NaCl, 3 mM MgCl , 0.1% Triton X-100) was added and incubated at 37 °C for 2 h. The cells in the lower layer were transferred to a new tube and digested by DNase I as described above. The formaldehyde cross-linking was reversed by incubating DNA at 65 °C overnight, which was followed by DNA purification and library preparation. Cells recovered from FFPE slides were resuspended in 150 μl of de-paraffinization solution (Qiagen, 1064343) and incubated at 56 °C for 3 min. Total RNA was extracted using an RNA extraction kit from (Qiagen, 73504), following the manufacturer’s instructions. After reverse transcription using an oligonucleotide dT primer, the mRNA expression levels of selected genes were analysed using the following gene-specific primers and probes from Applied Biosystems: HMGA2-Hs00171569_ml, TIAM1-Hs01021959_ml, TXNL1-Hs00355488_ml, PIP4K2A-Hs00178197_ml and GADPH-Hs99999905_ml. The RNA-seq libraries were generated according to established protocols and sequenced on HiSeq 2500 platforms. The tumour and adjacent normal cells from FFPE slides were recovered and resuspended in 100 μl of 1× TE + 0.1% SDS + 0.2 mg ml−1 proteinase K. Following incubation at 65 °C overnight, the genomic DNA was purified using phenol–chloroform extraction and ethanol precipitation. The genomic region containing the potential sequence variation was amplified by PCR using specific primers. The PCR products were then sequenced by Sanger sequencing. Forward primer, AAGCTAAATGAGCAAAATATTCCT; reverse primer, GGGAGGCTGAGGCAGTAGAATCG. Chromatin extracts were prepared from a human thyroid cell line (Nthy-ori 3-1 human Cell Line, from Sigma-Aldrich, 90011609). ChIP experiments were performed with p53 antibodies (Santa Cruz Biotechnology, sc-6243X) using established protocols1. The ChIP DNA was analysed using qPCR with the following primers: p53 positive forward primer, GTCATGCGATCTTGGCTCACT; reverse primer, CTTGGGAGGCTGAGGCAGTA; probe, CAACCTCCGCCTCCCGGGTTC. Control forward primer, CCCCATGCTGTTCTCGTGATA; reverse primer, GCAAAGGTGAATCAAGGCATCT; probe, TTTATAAGGTTCTCTTCC CCTTTCGCTGGG. Electrophoretic mobility shift assay (EMSA) experiments were performed using nuclear extracts of HeLa cells transfected with a p53 expression vector (provided by J. Huang). Briefly, the double-stranded oligonucleotide probes (wild-type p53 site, CACTCTGTTGCCCGGGCTAGTGTGCAGT; tumour p53 site, CACTCTGTTGCCCGGGCTACTGTGCAGT; p21 promoter p53 site, CAGGAACAAGTCAAGACATGTTCAGC) were synthesized and labelled with biotin using Biotin 3zeEnd DNA Labelling Kit (Thermo Scientific, 89818). The EMSA assays were conducted by using LightShift Chemiluminescent EMSA Kit (Thermo Scientific, 20148) according to the manufacturer’s instructions. To test the activity of the p53 binding sites to activate a reporter promoter, we cloned the wild type p53 binding motif, the motif with the G to C mutation and the p53 motif from the p21 promoter into the XhoI and BglII upstream of the basal cytomegalovirus promoter driving a luciferase reporter gene (provided by J. Huang). The constructs were transfected into Nthy-ori 3-1 human cell line cells for 2 days and the luciferase activity of whole-cell extracts was measured using a Dual-Luciferase Reporter Assay kit (Promega, E1960). The oligonucleotide sequences used in the reporter constructs were as follows: wild type p53 site, TCGAGCTGTTGCCCGGGCTAGTGTGA; tumour p53 site, TCGAGCTGTTGCCCGGGCTACTGTGA; p21 promoter p53 site, TCGAGGAACAAGTCAAGACATGTTCA. Data, reads mapping and filtering: in this study, we constructed a total of 38 scDNase-seq libraries including 8 NIH3T3 libraries (Supplementary Table 1), 18 ESC libraries (Supplementary Table 6) and 12 FFPE patient libraries (Supplementary Table 12). Among these libraries, there are 5 NIH3T3 single-cell scDNase-seq libraries and 14 ESC single-cell scDNase-seq libraries. We also prepared eight RNA-seq libraries using cells recovered from the FFPE tissue section slides of FTC 440 (Supplementary Table 12). In addition to the scDNase-seq and RNA-seq libraries prepared in this study, we integrated the histone modification ChIP-seq data of NIH3T3 from our previous study30. We also downloaded the DNase-seq data of NIH3T3 cells and ESCs from mouse ENCODE project31. Reads of DNase-seq/scDNase-seq/ChIP-seq were mapped to the mouse genome (mm9) or human genome (hg18) using Bowtie2 (ref. 32). Iterative alignment, in which the unmapped reads were trimmed 5 bp and were re-aligned until reads were less than 26 bp, were conducted for small cell number scDNase-seq libraries and single-cell scDNase-seq libraries. The reads with mapping quality (MAPQ) ≤ 10 or redundant reads that mapped to the same location with the same orientation were removed from further analysis in each library. The mappability of 1,000-cells scDNase-seq libraries to the mouse or human genome was about 40% whereas that of the single-cell scDNase-seq libraries was about 2% owing to non-specific amplification of carrier DNA. The tag density at each bin of 200 bp was calculated by normalizing the number of reads in the bin to the total number of reads in the library and the bedgraphs were uploaded to the UCSC Genome Browser. Peak calling for DNase-seq/scDNase-seq and correlation between different libraries: the DHSs in mouse ENCODE DNase-seq data and small cell number scDNase-seq data were identified using model-based analysis of ChIP-seq (MACS)33 by setting a P value to 1 × 10−5. The peaks identified in the ENCODE data were extended ±1 kb from the summit of the peak if the peak size was <2 kb and overlapping peaks were merged. Then the number of reads in each DHS for all DNase-seq and scDNase-seq libraries was counted. The tag density at each DHS was calculated by normalizing the number of reads in the DHS to the total number of reads in the library (possibility of a tag located on a base-pair per million reads). The Pearson product-moment correlation coefficient (r) of tag densities at genome-wide DHS between two libraries was calculated to indicate the correlation between different scDNase-seq libraries. For single-cell libraries, the reads out of the defined DHS regions were filtered and the number of reads in each 1,000-bp bin was counted to generate the single-cell heat map (Fig. 1b). Any DHS region in a single cell with a reads located in was treated as open access thus a DHS in this single cell. For the pooled five single cells, any DHS region with at least two reads located in was treated as the DHS in the pooled five single cells. The FDR of the DHS detected in single cells: in an NIH3T3 single-cell scDNase-seq library, the total number of observed DHSs and false positive (type I error) DHSs were denoted by N and N , respectively. On the other hand, any reads that located out of the DHSs detected in ENCODE data must have been caused by noise generated during library preparation. The noise level (σ) should be the total number of reads that located out of the DHS in ENCODE data dividing by total length of the regions that are not DHS. The number of false positive DHSs should be the genome-wide noise level (σ) multiplying by the total length of the DHS. Thus, the FDR should be the number of false positive DHSs dividing by all the detected DHSs in single cell: On the basis of this formula, we calculated the FDR for each NIH3T3 and ESC single-cell scDNase-seq library (Supplementary Tables 2 and 7). Differentially expressed genes and tissue-specific genes: the reads from RNA-seq libraries were mapped to the mouse genome (mm9) or human genome (hg18) using Bowtie2 (ref. 32). The gene expression level was measured by reads per kilobase per million mapped reads (RPKM) and number of reads in each gene. The cell-specific genes between ESC and NIH-3T3 were identified using EdgeR (FDR < 0.05; fold change > 1.5 or greater than two-thirds)34. We used the tissue specificity index τ (ref. 35) to measure the tissue specificity of each gene, which is defined as the heterogeneity of its expression level across all the tissues. Assuming there are n tissues, the expression level of a gene in the jth tissue is E(j) and the highest expression level of the gene across all tissues is E . Thus τ is calculated by The values of τ range from 0 to 1, with higher values indicating higher variation of expression across tissues and thus higher tissue specificity, whereas lower values indicate lower variation of expression across tissues. The genes with the lowest τ could be considered as housekeeping genes. In this study, we calculated τ on the basis of gene atlas data from bioGPS. The 2,000 genes with the highest τ and the 2,000 genes with the lowest τ were treated as the tissue-specific genes and housekeeping genes, respectively. The histone modification ChIP-seq data and peak calling: since the peaks of some histone marks such as H3K36me3 and H3K27me3 are very broad, we identified the tag-enriched peaks using SICER36, which takes advantage of the enrichment information from neighbouring bins to identify spatial clusters of signals that are unlikely to appear by chance. We set the window size to 200 bp and FDR = 0.01 for each histone modification ChIP-seq library, while we set the gap to 200 bp for H3K4me3, H3K9ac; 400 bp for H2A.Z; and 600 bp for the H3K4me1, H3K9me2, H3K27ac and H3K27me3. We calculated the tag densities of each active histone modification peak and identified whether the peak was a DHS in each single cell to find whether the enrichment of an active histone mark was correlated with the number of cells with DHS at the same locus. We calculated the tag densities of each single-cell scDNase-seq library at each DHS and examined whether a DHS co-occurred with these active histone modifications to find whether the chromatin accessibility in each single cell was correlated with the number of histone modifications in the same locus. Two peaks from different libraries were considered a co-occurrence if the overlapped region accounted for >10% of the length of a peak. Reads around promoters and subpeaks of super-enhancers: the RefSeq genes (mm9 and hg18) were downloaded from the UCSC Genome Browser database. The regions ±1 kb around the TSS were treated as promoters in this study. The number of scDNase-seq reads located in a promoter was used to measure the chromatin accessibility of the promoter. We searched the super-enhancer in NIH3T3 via ROSE15 on the basis of H3K27ac ChIP-seq and scDNase-seq data, respectively. We obtained a total of 275 high-confidence super-enhancers in NIH3T3 by identifying super-enhancers shown both in H3K27ac and in DNase-seq data. In addition, the 231 super-enhancers in ESCs reported in ref. 15 were used in this study. Subpeaks in super-enhancers were identified by MACS33 and average read densities around these subpeaks of super-enhancers were calculated. Single-cell-specific DHSs and gene set enrichment analysis: the number of reads located each DHS detected in ENCODE data in each NIH3T3 cell and ESC was counted. To examine whether the chromatin accessibility between NIH3T3 cells and ESCs was significantly different, a Wilcoxon signed-rank test was performed on the number of reads in the 5 NIH3T3 cells and 14 ESCs at each DHS. A DHS was active (indicated by 1) in a single cell if there was one or more than one reads located in the DHS region in the cell, while the alternative was not active (indicated by 0). Fisher’s exact test on each locus was performed on the number of cells with active DHSs and the number of cells without active DHSs between the 5 NIH3T3 cells and 14 ESCs. The DHSs with P < 0.05 both by Wilcoxon test and by Fisher’s test were treated as cell-type specific. Finally, we identified 1,735 single-cell NIH3T3-specific DHSs and 2,180 single-cell ESC-specific DHSs. We used gene set enrichment analysis37 to determine whether the genes in the vicinity of the single-cell-specific DHSs showed statistically significant differences between NIH3T3 cells and ESCs on the basis of the gene expression data. Gene ontology of single-cell NIH3T3-specific and ESC-specific DHSs: to predict the function of single-cell NIH3T3-specific or ESC-specific DHSs, we performed gene ontology analysis using GREAT38 with the 1,735 NIH3T3-specific and 2,180 ESC-specific DHSs. It is clear that the single-cell ESC-specific DHSs are enriched with stem cell development and differentiation genes, and the single-cell NIH3T3-specific DHSs are enriched with genes with different functions (Extended Data Fig. 6g, h). These results indicate that the ESC-specific and NIH3T3-specific DHSs identified in the single-cell scDNase-seq libraries predict important enhancers critical for tissue-specific gene expression. Identifying tumour-specific mutation: we generated scDNase-seq libraries using tumour or their neighbouring cells recovered from FFPE tissue section slides. The sequence reads were mapped by Bowtie2 (ref. 32)32. to the human reference genome (hg18). The paired reads with distance < 500 bp were kept if paired-end sequencing was performed. Then reads with MAPQ < 20 and possible duplication were removed by SAMtools39. Variation calling on each normal–tumour pair was conducted using SAMtools mpileup, with diploid model, MAPQ ≥ 20 and base alignment quality (BAQ) ≥ 30. The variations that only normal and tumour show different genotypes were kept. Then the low-quality variations were filtered (query quality (QUAL) < 20, mapping quality (MQ) < 20, phred probability of all samples being the same (FQ) < 0, variant distance bias (VDB < 0.01 and minor allele < 3). We obtained 31 variation candidates in FTC 440 (Supplementary Table 13), many of them located on the predicted transcription-factor binding motifs. Tumour- and normal-cell-specific DHS: the genome-wide DHSs were obtained by peak calling of the normal cell and tumour cell scDNase-seq libraries, respectively. The DHSs in normal cells and tumour cells were pooled, and reads in each library among the pooled DHSs were counted. The normal- and tumour-cell-specific DHSs were identified using EdgeR.
Fermat proof prize Andrew Wiles has received the 2016 Abel Prize for mathematics for his solution to Fermat’s last theorem, the Norwegian Academy of Science and Letters announced on 15 March. The problem had stumped some of the world’s greatest minds for three and a half centuries. Wiles, a number theorist now at the University of Oxford, UK, will receive 6 million kroner (US$700,000) for his 1994 proof showing that there cannot be any positive whole numbers x, y and z such that xn + yn = zn, if n is greater than 2. See go.nature.com/yf1nxj for more. Famous killer whale nears end of life Tilikum, a killer whale (Orcinus orca) at SeaWorld in Orlando, Florida, has an incurable lung infection, the theme park’s veterinary team has announced. In February 2010, Tilikum dragged his trainer Dawn Brancheau into the pool and killed her. The whale was also involved in two deaths in the 1990s, and the story of his life in captivity was told in the controversial 2013 documentary film Blackfish. SeaWorld bought Tilikum in 1983; he is thought to be 35 years old. The species’ life expectancy in captivity versus that in the wild is still debated by scientists. AlphaGo victorious The world’s leading Go player, South Korea’s Lee Sedol, lost his final match in Seoul against Google DeepMind’s AlphaGo machine on 15 March. The tightly fought game brought the best-of-five competition to an end with four wins for the computer versus one for the human player. Sedol came back from three consecutive losses to beat the artificial-intelligence system in the fourth match, but ultimately missed out on the US$1-million prize. Go originated more than 2,500 years ago in China and involves placing black and white counters on a board. See page 284 for more. Brexit warning Physicist Stephen Hawking is one of more than 150 scientists, mathematicians, economists and engineers at the University of Cambridge, UK, who warn of a disaster for the nation’s science if Britain exits the European Union (known as Brexit). A referendum to be held on 23 June will ask whether the country should leave the EU. In a 10 March letter to The Times, organized by protein scientist Alan Fersht, the group argues that the free movement of workers between EU countries helps in the recruitment of high-quality researchers to the United Kingdom. The letter’s signatories are all fellows of the Royal Society in London. Zika meeting With the Zika virus still spreading rapidly across the Americas, the World Health Organization (WHO) in Geneva held an emergency meeting on mosquito control on 14–15 March. The WHO’s Vector Control Advisory Group intends to review evidence to support new and innovative techniques for combating the Aedes aegypti mosquitoes that transmit Zika virus, along with dengue and Chikungunya viruses. These techniques include deploying mosquitoes that have been made infertile through genetic modification or irradiation. Infrastructure map The European Commission has published its latest wish list of the research-infrastructure projects that it considers most deserving of continent-wide support. The European Strategy Forum on Research Infrastructures road map, released on 10 March, details 21 facilities across all scientific areas to help national governments to prioritize how they spend infrastructure money, and to encourage them to share costs and responsibilities. New facilities listed in the 2016 road map include two in environmental sciences and one in health and food sciences, as well as solar and neutrino telescopes and an infrastructure for scientific research into cultural heritage. Minister keeps title German defence minister Ursula von der Leyen, who was accused in September 2015 of plagiarism in her medical dissertation in obstetrics, will not lose the title of doctor or her job. The senate of Hanover Medical School, which awarded the title in 1990, announced on 9 March that its formal investigation revealed that some passages in von der Leyen’s dissertation were copied from original sources. But these were mostly in the introduction, it said, and the main body of research was original and valid. Since 2011, two German federal ministers have lost their titles and government posts to plagiarism charges. Call to save bees The US Government Accountability Office (GAO) says that US regulatory bodies need to do more to protect bee populations. In a report made public on 11 March, the GAO called on the US Department of Agriculture (USDA) to work more closely with other agencies to protect bee health. The report says that although the USDA has upped efforts to monitor honeybee colonies managed by beekeepers, it does not coordinate the monitoring of wild, native bees. The report also recommends that the Environmental Protection Agency identifies the mixtures of pesticides most commonly used by farmers. Gene data shared Researchers and the public can now access a database of anonymized genetic information from 10,000 people with hereditary breast or ovarian cancer. The database, called AmbryShare, was launched on 8 March by Ambry Genetics, a genetic-testing company in Aliso Viejo, California — making Ambry the first private company to release its customers’ information for free. The Broad Institute of MIT and Harvard in Cambridge, Massachusetts, has an open-access database of more than 60,000 genomes collected from the public, but AmbryShare’s data currently focus on specific diseases. Ambry hopes to release up to 200,000 aggregated genomes per year from people with various conditions. India vaccine fight The medical charity Médecins Sans Frontières (MSF) is challenging pharmaceutical company Pfizer’s application for a patent in India on pneumonia vaccine PCV13, marketed as Prevenar 13 in India. MSF says that it wants to allow other manufacturers to make the vaccine, and lower its cost. The 11 March challenge asserts that the method that Pfizer is trying to patent is too obvious to deserve a patent under Indian law. Pfizer is reported as saying that the complexity of the vaccine justifies the price. In partnership with the vaccine alliance GAVI, Pfizer has reduced the price of Prevenar since 2013. Mosquito trial A proposed field trial of genetically modified mosquitoes in the Florida Keys poses no threat to human health or the environment, the US Food and Drug Administration has determined. Members of the public have 30 days to submit comments on the draft assessment, which was released on 11 March. The Aedes aegypti mosquitoes developed by Oxitec of Oxford, UK, are engineered to produce short-lived young to temporarily reduce mosquito populations and combat diseases that they carry. The project has received increased attention from the media and politicians amid concerns about the spread of Zika virus. The level of atmospheric carbon dioxide at the Mauna Loa Observatory in Hawaii rose by 3.05 parts per million (p.p.m.) in 2015 — the largest annual increase since records began 56 years ago, says the US National Oceanic and Atmospheric Administration. After correcting for seasonal swings from plant-growth cycles in the Northern Hemisphere, the average CO concentration in 2015 was 400.83 p.p.m. — a 43% rise compared to the CO level of around 280 p.p.m. that existed during the pre-industrial era. 10 Consecutive months in which the global monthly temperature record has been broken. February’s temperature was 1.35 °C above average for the month. A strong El Niño weather system has contributed to the record-breaking run. Source: NOAA 17–18 March Commercializing 3D printing for biological applications is discussed at the second Tissue Engineering, Biofabrication & 3D-Bioprinting in Life Sciences conference in Boston, Massachusetts. go.nature.com/rggrat 21–23 March NASA holds a meeting in Washington DC to develop its technology road maps. go.nature.com/dhmq2e 21–25 March The annual Lunar and Planetary Science Conference convenes in The Woodlands, Texas. go.nature.com/qpnoxd
Minion L.E.,Dignity Health St Josephs Hospital And Medical Center |
Dolinsky J.S.,Ambry Genetics |
Chase D.M.,Arizona Cancer Center |
Dunlop C.L.,Ambry Genetics |
And 3 more authors.
Gynecologic Oncology | Year: 2015
Objective. Genetic predisposition to ovarian cancer is well documented.With the advent of next generation sequencing, hereditary panel testing provides an efficientmethod for evaluating multiple genes simultaneously. Therefore, we sought to investigate the contribution of 19 genes identified in the literature as increasing the risk of hereditary breast and ovarian cancer (HBOC) in a BRCA1 and BRCA2 negative population of patients with a personal history of breast and/or ovarian cancer by means of a hereditary cancer panel. Methods. Subjects were referred for multi-gene panel testing between February 2012 and March 2014. Clinical data was ascertained from requisition forms. The incidence of pathogenic mutations (including likely pathogenic), and variant of unknown significance were then calculated for each gene and/or patient cohort. Results. In this cohort of 911 subjects, panel testing identified 67 mutations.With 7.4% of subjects harboring a mutation on this multi-gene panel, the diagnostic yield was increased, compared to testing for BRCA1 and BRCA2 mutations alone. In the ovarian cancer probands, the most frequentlymutated genes were BRIP1 (n=8; 1.72%) and MSH6 (n = 6; 1.29%). In the breast cancer probands, mutations were most commonly observed in CHEK2 (n= 9; 2.54%), ATM (n= 3; 0.85%), and TP53 (n= 3; 0.85%). Conclusions. Although further studies are needed to clarify the exact management of patientswith amutation in each gene, this study highlights information that can be captured with panel testing and provides support for incorporation of panel testing into clinical practice. © 2015 The Authors. Published by Elsevier Inc.