In this work, we present the Genome Modeling System (GMS), an analysis information management system capable of executing automated genome analysis pipelines at a massive scale. The GMS framework provides detailed tracking of samples and data coupled with reliable and repeatable analysis pipelines. The GMS also serves as a platform for bioinformatics development, allowing a large team to collaborate on data analysis, or an individual researcher to leverage the work of others effectively within its data management system. Rather than separating ad-hoc analysis from rigorous, reproducible pipelines, the GMS promotes systematic integration between the two. As a demonstration of the GMS, we performed an integrated analysis of whole genome, exome and transcriptome sequencing data from a breast cancer cell line (HCC1395) and matched lymphoblastoid line (HCC1395BL). These data are available for users to test the software, complete tutorials and develop novel GMS pipeline configurations. The GMS is available at https://github.com/genome/gms.
Despite the success of genome-wide association studies (GWAS) in detecting a large number of loci for complex phenotypes such as rheumatoid arthritis (RA) susceptibility, the lack of information on the causal genes leaves important challenges to interpret GWAS results in the context of the disease biology. Here, we genetically fine-map the RA risk locus at 19p13 to define causal variants, and explore the pleiotropic effects of these same variants in other complex traits. First, we combined Immunochip dense genotyping (n = 23,092 case/control samples), Exomechip genotyping (n = 18,409 case/control samples) and targeted exon-sequencing (n = 2,236 case/controls samples) to demonstrate that three protein-coding variants in TYK2 (tyrosine kinase 2) independently protect against RA: P1104A (rs34536443, OR = 0.66, P = 2.3x10-21), A928V (rs35018800, OR = 0.53, P = 1.2x10-9), and I684S (rs12720356, OR = 0.86, P = 4.6x10-7). Second, we show that the same three TYK2 variants protect against systemic lupus erythematosus (SLE, Pomnibus = 6x10-18), and provide suggestive evidence that two of the TYK2 variants (P1104A and A928V) may also protect against inflammatory bowel disease (IBD; Pomnibus = 0.005). Finally, in a phenome-wide association study (PheWAS) assessing >500 phenotypes using electronic medical records (EMR) in >29,000 subjects, we found no convincing evidence for association of P1104A and A928V with complex phenotypes other than autoimmune diseases such as RA, SLE and IBD. Together, our results demonstrate the role of TYK2 in the pathogenesis of RA, SLE and IBD, and provide supporting evidence for TYK2 as a promising drug target for the treatment of autoimmune diseases.
Detecting somatic single nucleotide variants (SNVs) is an essential component of cancer research with next generation sequencing data. This protocol describes how to run the SomaticSniper somatic SNV detector and then filter the output to eliminate most false positives. It also includes support protocols detailing the compilation of the software.
next-generation sequencing; somatic variant calling; somaticsniper
The identification of small sequence variants remains a challenging but critical step in the analysis of next-generation sequencing data. Our variant calling tool, VarScan 2, employs heuristic and statistic thresholds based on user-defined criteria to call variants using SAMtools mpileup data as input. Here, we provide guidelines for generating that input, and describe protocols for using VarScan 2 to (1) identify germline variants in individual samples; (2) call somatic mutations, copy number alterations, and LOH events in tumor-normal pairs; and (3) identify germline variants, de novo mutations, and Mendelian inheritance errors in family trios. Further, we describe a strategy for variant filtering that removes likely false positives associated with common sequencing- and alignment-related artifacts.
variant calling; mutation detection; trio calling; snvs; indels; varscan 2; next-generation sequencing
Genomics is a relatively new scientific discipline, having DNA sequencing as its core technology. As technology has improved the cost and scale of genome characterization over sequencing’s 40-year history, the scope of inquiry has commensurately broadened. Massively parallel sequencing has proven revolutionary, shifting the paradigm of genomics to address biological questions at a genome-wide scale. Sequencing now empowers clinical diagnostics and other aspects of medical care, including disease risk, therapeutic identification, and prenatal testing. This Review explores the current state of genomics in the massively parallel sequencing era.
We report the first large-scale exome-wide analysis of the combined germline-somatic landscape in ovarian cancer. Here we analyze germline and somatic alterations in 429 ovarian carcinoma cases and 557 controls. We identify 3,635 high confidence, rare truncation and 22,953 missense variants with predicted functional impact. We find germline truncation variants and large deletions across Fanconi pathway genes in 20% of cases. Enrichment of rare truncations is shown in BRCA1, BRCA2, and PALB2. Additionally, we observe germline truncation variants in genes not previously associated with ovarian cancer susceptibility (NF1, MAP3K4, CDKN2B, and MLL3). Evidence for loss of heterozygosity was found in 100% and 76% of cases with germline BRCA1 and BRCA2 truncations respectively. Germline-somatic interaction analysis combined with extensive bioinformatics annotation identifies 237 candidate functional germline truncation and missense variants, including 2 pathogenic BRCA1 and 1 TP53 deleterious variants. Finally, integrated analyses of germline and somatic variants identify significantly altered pathways, including the Fanconi, MAPK, and MLL pathways.
The Drug-Gene Interaction database (DGIdb) mines existing resources that generate hypotheses about how mutated genes might be targeted therapeutically or prioritized for drug development. It provides an interface for searching lists of genes against a compendium of drug-gene interactions and potentially druggable genes. DGIdb can be accessed at dgidb.org.
Macular degeneration is a common cause of blindness in the elderly. To identify rare coding variants associated with a large increase in risk of age-related macular degeneration (AMD), we sequenced 2,335 cases and 789 controls in 10 candidate loci (57 genes). To increase power, we augmented our control set with ancestry-matched exome sequenced controls. An analysis of coding variation in 2,268 AMD cases and 2,268 ancestry matched controls revealed two large-effect rare variants; previously described R1210C in the CFH gene (fcase = 0.51%, fcontrol = 0.02%, OR = 23.11), and newly identified K155Q in the C3 gene (fcase = 1.06%, fcontrol = 0.39%, OR = 2.68). The variants suggest decreased inhibition of C3 by Factor H, resulting in increased activation of the alternative complement pathway, as a key component of disease biology.
Genome-wide association studies (GWAS) have identified >500 common variants associated with quantitative metabolic traits, but in aggregate such variants explain at most 20–30% of the heritable component of population variation in these traits. To further investigate the impact of genotypic variation on metabolic traits, we conducted re-sequencing studies in >6,000 members of a Finnish population cohort (The Northern Finland Birth Cohort of 1966 [NFBC]) and a type 2 diabetes case-control sample (The Finland-United States Investigation of NIDDM Genetics [FUSION] study). By sequencing the coding sequence and 5′ and 3′ untranslated regions of 78 genes at 17 GWAS loci associated with one or more of six metabolic traits (serum levels of fasting HDL-C, LDL-C, total cholesterol, triglycerides, plasma glucose, and insulin), and conducting both single-variant and gene-level association tests, we obtained a more complete understanding of phenotype-genotype associations at eight of these loci. At all eight of these loci, the identification of new associations provides significant evidence for multiple genetic signals to one or more phenotypes, and at two loci, in the genes ABCA1 and CETP, we found significant gene-level evidence of association to non-synonymous variants with MAF<1%. Additionally, two potentially deleterious variants that demonstrated significant associations (rs138726309, a missense variant in G6PC2, and rs28933094, a missense variant in LIPC) were considerably more common in these Finnish samples than in European reference populations, supporting our prior hypothesis that deleterious variants could attain high frequencies in this isolated population, likely due to the effects of population bottlenecks. Our results highlight the value of large, well-phenotyped samples for rare-variant association analysis, and the challenge of evaluating the phenotypic impact of such variants.
Abnormal serum levels of various metabolites, including measures relevant to cholesterol, other fats, and sugars, are known to be risk factors for cardiovascular disease and type 2 diabetes. Identification of the genes that play a role in generating such abnormalities could advance the development of new treatment and prevention strategies for these disorders. Investigations of common genetic variants carried out in large sets of research subjects have successfully pinpointed such genes within many regions of the human genome. However, these studies often have not led to the identification of the specific genetic variations affecting metabolic traits. To attempt to detect such causal variations, we sequenced genes in 17 genomic regions implicated in metabolic traits in >6,000 people from Finland. By conducting statistical analyses relating specific variations (individually and grouped by gene) to the measures for these metabolic traits observed in the study subjects, we added to our understanding of how genotypes affect these traits. Our findings support a long-held hypothesis that the unique history of the Finnish population provides important advantages for analyzing the relationship between genetic variations and biomedically important traits.
Acute promyelocytic leukemia (APL) is a subtype of acute myeloid leukemia (AML). It is characterized by the t(15;17)(q22;q11.2) chromosomal translocation that creates the promyelocytic leukemia–retinoic acid receptor α (PML-RARA) fusion oncogene. Although this fusion oncogene is known to initiate APL in mice, other cooperating mutations, as yet ill defined, are important for disease pathogenesis. To identify these, we used a mouse model of APL, whereby PML-RARA expressed in myeloid cells leads to a myeloproliferative disease that ultimately evolves into APL. Sequencing of a mouse APL genome revealed 3 somatic, nonsynonymous mutations relevant to APL pathogenesis, of which 1 (Jak1 V657F) was found to be recurrent in other affected mice. This mutation was identical to the JAK1 V658F mutation previously found in human APL and acute lymphoblastic leukemia samples. Further analysis showed that JAK1 V658F cooperated in vivo with PML-RARA, causing a rapidly fatal leukemia in mice. We also discovered a somatic 150-kb deletion involving the lysine (K)-specific demethylase 6A (Kdm6a, also known as Utx) gene, in the mouse APL genome. Similar deletions were observed in 3 out of 14 additional mouse APL samples and 1 out of 150 human AML samples. In conclusion, whole genome sequencing of mouse cancer genomes can provide an unbiased and comprehensive approach for discovering functionally relevant mutations that are also present in human leukemias.
Most mutations in cancer genomes are thought to be acquired after the initiating event, which may cause genomic instability, driving clonal evolution. However, for acute myeloid leukemia (AML), normal karyotypes are common, and genomic instability is unusual. To better understand clonal evolution in AML, we sequenced the genomes of AML samples with a known initiating event (PML-RARA) vs. normal karyotype AML samples, and the exomes of hematopoietic stem/progenitor cells (HSPCs) from healthy people. Collectively, the data suggest that most of the mutations found in AML genomes are actually random events that occurred in HSPCs before they acquired the initiating mutation; the mutational history of that cell is “captured” as the clone expands. In many cases, only one or two additional, cooperating mutations are needed to generate the malignant founding clone. Cells from the founding clone can acquire additional cooperating mutations, yielding subclones that can contribute to disease progression and/or relapse.
The emergence of next-generation sequencing (NGS) technologies offers an incredible opportunity to comprehensively study DNA sequence variation in human genomes. Commercially available platforms from Roche (454), Illumina (Genome Analyzer and Hiseq 2000), and Applied Biosystems (SOLiD) have the capability to completely sequence individual genomes to high levels of coverage. NGS data is particularly advantageous for the study of structural variation (SV) because it offers the sensitivity to detect variants of various sizes and types, as well as the precision to characterize their breakpoints at base pair resolution. In this chapter, we present methods and software algorithms that have been developed to detect SVs and copy number changes using massively parallel sequencing data. We describe visualization and de novo assembly strategies for characterizing SV breakpoints and removing false positives.
Next-generation sequencing; Paired-end sequencing; 454; Illumina; Solexa; Abi solid; Insertions; Deletions; Duplications; Inversions; Translocations; Indels; Copy number variants
Detection and characterization of genomic structural variation are important for understanding the landscape of genetic variation in human populations and in complex diseases such as cancer. Recent studies demonstrate the feasibility of detecting structural variation using next-generation, short-insert, paired-end sequencing reads. However, the utility of these reads is not entirely clear, nor are the analysis methods under which accurate detection can be achieved. The algorithm BreakDancer predicts a wide variety of structural variants including indels, inversions, and translocations. We examined BreakDancer's performance in simulation, comparison with other methods, analysis of an acute myeloid leukemia sample, and the 1,000 Genomes trio individuals. We found that it substantially improved the detection of small and intermediate size indels from 10 bp to 1 Mbp that are difficult to detect via a single conventional approach.
To assess the genetic consequences of induced Pluripotent Stem Cell (iPSC) reprogramming, we sequenced the genomes of ten murine iPSC clones derived from three independent reprogramming experiments, and compared them to their parental cell genomes. We detected hundreds of single nucleotide variants (SNVs) in every clone, with an average of 11 in coding regions. In two experiments, all SNVs were unique for each clone and did not cluster in pathways, but in the third, all four iPSC clones contained 157 shared genetic variants, which could also be detected in rare cells (<1 in 500) within the parental MEF pool. This data suggests that most of the genetic variation in iPSC clones is not caused by reprogramming per se, but is rather a consequence of cloning individual cells, which “captures” their mutational history. These findings have implications for the development and therapeutic use of cells that are reprogrammed by any method.
Motivation: The sequencing of tumors and their matched normals is frequently used to study the genetic composition of cancer. Despite this fact, there remains a dearth of available software tools designed to compare sequences in pairs of samples and identify sites that are likely to be unique to one sample.
Results: In this article, we describe the mathematical basis of our SomaticSniper software for comparing tumor and normal pairs. We estimate its sensitivity and precision, and present several common sources of error resulting in miscalls.
Availability and implementation: Binaries are freely available for download at http://gmt.genome.wustl.edu/somatic-sniper/current/, implemented in C and supported on Linux and Mac OS X.
Contact: firstname.lastname@example.org; email@example.com
Supplementary information: Supplementary data are available at Bioinformatics online.
The myelodysplastic syndromes are a group of hematologic disorders that often evolve into secondary acute myeloid leukemia (AML). The genetic changes that underlie progression from the myelodysplastic syndromes to secondary AML are not well understood.
We performed whole-genome sequencing of seven paired samples of skin and bone marrow in seven subjects with secondary AML to identify somatic mutations specific to secondary AML. We then genotyped a bone marrow sample obtained during the antecedent myelodysplastic-syndrome stage from each subject to determine the presence or absence of the specific somatic mutations. We identified recurrent mutations in coding genes and defined the clonal architecture of each pair of samples from the myelodysplastic-syndrome stage and the secondary-AML stage, using the allele burden of hundreds of mutations.
Approximately 85% of bone marrow cells were clonal in the myelodysplastic-syndrome and secondary-AML samples, regardless of the myeloblast count. The secondary-AML samples contained mutations in 11 recurrently mutated genes, including 4 genes that have not been previously implicated in the myelodysplastic syndromes or AML. In every case, progression to acute leukemia was defined by the persistence of an antecedent founding clone containing 182 to 660 somatic mutations and the outgrowth or emergence of at least one subclone, harboring dozens to hundreds of new mutations. All founding clones and subclones contained at least one mutation in a coding gene.
Nearly all the bone marrow cells in patients with myelodysplastic syndromes and secondary AML are clonally derived. Genetic evolution of secondary AML is a dynamic process shaped by multiple cycles of mutation acquisition and clonal selection. Recurrent gene mutations are found in both founding clones and daughter subclones. (Funded by the National Institutes of Health and others.)
Most patients with acute myeloid leukemia (AML) die from progressive disease after relapse, which is associated with clonal evolution at the cytogenetic level1,2. To determine the mutational spectrum associated with relapse, we sequenced the primary tumor and relapse genomes from 8 AML patients, and validated hundreds of somatic mutations using deep sequencing; this allowed us to precisely define clonality and clonal evolution patterns at relapse. Besides discovering novel, recurrently mutated genes (e.g. WAC, SMC3, DIS3, DDX41, and DAXX) in AML, we found two major clonal evolution patterns during AML relapse: 1) the founding clone in the primary tumor gained mutations and evolved into the relapse clone, or 2) a subclone of the founding clone survived initial therapy, gained additional mutations, and expanded at relapse. In all cases, chemotherapy failed to eradicate the founding clone. The comparison of relapse-specific vs. primary tumor mutations in all 8 cases revealed an increase in transversions, probably due to DNA damage caused by cytotoxic chemotherapy. These data demonstrate that AML relapse is associated with the addition of new mutations and clonal evolution, which is shaped in part by the chemotherapy that the patients receive to establish and maintain remissions.
Myelodysplastic syndromes (MDS) are hematopoietic stem cell disorders that often progress to chemotherapy-resistant secondary acute myeloid leukemia (sAML). We used whole genome sequencing to perform an unbiased comprehensive screen to discover all the somatic mutations in a sAML sample and genotyped these loci in the matched MDS sample. Here we show that a missense mutation affecting the serine at codon 34 (S34) in U2AF1 was recurrently mutated in 13/150 (8.7%) de novo MDS patients, with suggestive evidence of an associated increased risk of progression to sAML. U2AF1 is a U2 auxiliary factor protein that recognizes the AG splice acceptor dinucleotide at the 3′ end of introns and mutations are located in highly conserved zinc fingers in U2AF11,2. Mutant U2AF1 promotes enhanced splicing and exon skipping in reporter assays in vitro. This novel, recurrent mutation in U2AF1 implicates altered pre-mRNA splicing as a potential mechanism for MDS pathogenesis.
The full complement of DNA mutations that are responsible for the pathogenesis of acute myeloid leukemia (AML) is not yet known.
We used massively parallel DNA sequencing to obtain a very high level of coverage (approximately 98%) of a primary, cytogenetically normal, de novo genome for AML with minimal maturation (AML-M1) and a matched normal skin genome.
We identified 12 acquired (somatic) mutations within the coding sequences of genes and 52 somatic point mutations in conserved or regulatory portions of the genome. All mutations appeared to be heterozygous and present in nearly all cells in the tumor sample. Four of the 64 mutations occurred in at least 1 additional AML sample in 188 samples that were tested. Mutations in NRAS and NPM1 had been identified previously in patients with AML, but two other mutations had not been identified. One of these mutations, in the IDH1 gene, was present in 15 of 187 additional AML genomes tested and was strongly associated with normal cytogenetic status; it was present in 13 of 80 cytogenetically normal samples (16%). The other was a nongenic mutation in a genomic region with regulatory potential and conservation in higher mammals; we detected it in one additional AML tumor. The AML genome that we sequenced contains approximately 750 point mutations, of which only a small fraction are likely to be relevant to pathogenesis.
By comparing the sequences of tumor and skin genomes of a patient with AML-M1, we have identified recurring mutations that may be relevant for pathogenesis.
The genetic alterations responsible for an adverse outcome in most patients with acute myeloid leukemia (AML) are unknown.
Using massively parallel DNA sequencing, we identified a somatic mutation in DNMT3A, encoding a DNA methyltransferase, in the genome of cells from a patient with AML with a normal karyotype. We sequenced the exons of DNMT3A in 280 additional patients with de novo AML to define recurring mutations.
A total of 62 of 281 patients (22.1%) had mutations in DNMT3A that were predicted to affect translation. We identified 18 different missense mutations, the most common of which was predicted to affect amino acid R882 (in 37 patients). We also identified six frameshift, six nonsense, and three splice-site mutations and a 1.5-Mbp deletion encompassing DNMT3A. These mutations were highly enriched in the group of patients with an intermediate-risk cytogenetic profile (56 of 166 patients, or 33.7%) but were absent in all 79 patients with a favorable-risk cytogenetic profile (P<0.001 for both comparisons). The median overall survival among patients with DNMT3A mutations was significantly shorter than that among patients without such mutations (12.3 months vs. 41.1 months, P<0.001). DNMT3A mutations were associated with adverse outcomes among patients with an intermediate-risk cytogenetic profile or FLT3 mutations, regardless of age, and were independently associated with a poor outcome in Cox proportional-hazards analysis.
DNMT3A mutations are highly recurrent in patients with de novo AML with an intermediate-risk cytogenetic profile and are independently associated with a poor outcome. (Funded by the National Institutes of Health and others.)
Whole genome sequencing (WGS) is becoming increasingly available for research purposes, but it has not yet been routinely used for clinical diagnosis.
To determine whether whole genome sequencing can identify cryptic, actionable mutations in a clinically relevant time frame.
Design, Setting, and Patient
We were referred a difficult diagnostic case of acute promyelocytic leukemia with no pathogenic X-RARA fusion identified by routine metaphase cytogenetics or interphase FISH. The patient was enrolled in an IRB approved protocol, with consent specifically tailored to the implications of whole genome sequencing. The protocol employs a ‘movable firewall,’ which maintains patient anonymity within the entire research team, but allows the research team to communicate medically relevant information to the treating physician.
Main Outcome Measure
Clinical relevance of whole genome sequencing and time to communicate validated results to the treating physician.
Massively parallel paired-end sequencing allowed us to identify a cytogenetically cryptic event: 77 kilobases from chromosome 15 was inserted en bloc into the second intron of the RARA gene on chromosome 17, resulting in a classic bcr3 PML-RARA fusion gene. RT-PCR subsequently validated the expression of the fusion transcript. Novel FISH probes identified two additional cases of t(15;17)-negative acute promyelocytic leukemia that had cytogenetically invisible insertions. Whole genome sequencing and validation were completed in seven weeks, and changed the treatment plan for the patient.
Whole genome sequencing can identify cytogenetically invisible oncogenes in a clinically relevant timeframe.
Motivation: DNA copy number aberration (CNA) is a hallmark of genomic abnormality in tumor cells. Recurrent CNA (RCNA) occurs in multiple cancer samples across the same chromosomal region and has greater implication in tumorigenesis. Current commonly used methods for RCNA identification require CNA calling for individual samples before cross-sample analysis. This two-step strategy may result in a heavy computational burden, as well as a loss of the overall statistical power due to segmentation and discretization of individual sample's data. We propose a population-based approach for RCNA detection with no need of single-sample analysis, which is statistically powerful, computationally efficient and particularly suitable for high-resolution and large-population studies.
Results: Our approach, correlation matrix diagonal segmentation (CMDS), identifies RCNAs based on a between-chromosomal-site correlation analysis. Directly using the raw intensity ratio data from all samples and adopting a diagonal transformation strategy, CMDS substantially reduces computational burden and can obtain results very quickly from large datasets. Our simulation indicates that the statistical power of CMDS is higher than that of single-sample CNA calling based two-step approaches. We applied CMDS to two real datasets of lung cancer and brain cancer from Affymetrix and Illumina array platforms, respectively, and successfully identified known regions of CNA associated with EGFR, KRAS and other important oncogenes. CMDS provides a fast, powerful and easily implemented tool for the RCNA analysis of large-scale data from cancer genomes.
Availability: The R and C programs implementing our method are available at https://dsgweb.wustl.edu/qunyuan/software/cmds.
Supplementary information: Supplementary data are available at Bioinformatics online.
Correct patterning of cells within an epithelium is key to establishing their normal function. However, the precise mechanisms by which individual cells arrive at their final developmental niche remains poorly understood. We developed an optimized system for imaging the developing Drosophila retina, an ideal tissue for the study of cell positioning. Using this technique, we characterized the cellular dynamics of developing wild-type pupal retinas. We also analyzed two mutants affecting eye patterning and demonstrate that cells mutant for Notch or Roughest signaling were aberrantly dynamic in their cell movements. Finally, we establish a role for the adherens junction regulator P120-Catenin in retinal patterning through its regulation of normal adherens junction integrity. Our results indicate a requirement for P120-Catenin in the developing retina, the first reported developmental function of this protein in the epithelia of lower metazoa. Based upon our live visualization of the P120-Catenin mutant as well as genetic data, we conclude that P120-Catenin is acting to stabilize E-cadherin and adherens junction integrity during eye development.
Live visualization; p120Catenin; Drosophila; Retina; Tissue patterning
The correct organization of cells within an epithelium is essential for proper tissue and organ morphogenesis. The role of Decapentaplegic/Bone morphogenetic protein (Dpp/BMP) signaling in cellular morphogenesis during epithelial development is poorly understood. In this paper, we used the developing Drosophila pupal retina – looking specifically at the reorganization of glial-like support cells that lie between the retinal ommatidia – to better understand the role of Dpp signaling during epithelial patterning. Our results indicate that Dpp pathway activity is tightly regulated across time in the pupal retina and that epithelial cells in this tissue require Dpp signaling to achieve their correct shape and position within the ommatidial hexagon. These results point to the Dpp pathway as a third component and functional link between two adhesion systems, Hibris-Roughest and DE-cadherin. A balanced interplay between these three systems is essential for epithelial patterning during morphogenesis of the pupal retina. Importantly, we identify a similar functional connection between Dpp activity and DE-cadherin and Rho1 during cell fate determination in the wing, suggesting a broader link between Dpp function and junctional integrity during epithelial development.
Adhesion; BMP; Dpp; Epithelia; Patterning
Massively parallel DNA sequencing technologies provide an unprecedented ability to screen entire genomes for genetic changes associated with tumor progression. Here we describe the genomic analyses of four DNA samples from an African-American patient with basal-like breast cancer: peripheral blood, the primary tumor, a brain metastasis, and a xenograft derived from the primary tumor. The metastasis contained two de novo mutations and a large deletion not present in the primary tumor, and was significantly enriched for 20 shared mutations. The xenograft retained all primary tumor mutations, and displayed a mutation enrichment pattern that paralleled the metastasis (16 of 20 genes). Two overlapping large deletions, encompassing CTNNA1, were present in all three tumor samples. The differential mutation frequencies and structural variation patterns in metastasis and xenograft compared to the primary tumor suggest that secondary tumors may arise from a minority of cells within the primary.