Whole-genome Sequencing Reads Research Articles

Abstract LB-051: Tumor heterogeneity and dissemination in breast cancer: Deep sequencing of single disseminated cells from bone marrow compared to primary tumor and lymph node metastases

Abstract Metastasis is the main cause of death amongst breast cancer patients. Our knowledge of the metastatic cascade and how to inhibit it is limited. Here we dissect the genetic profile of multiple single disseminated tumor cells (DTCs) taken at various time points after diagnosis, and compare them to their matched primary tumors and lymph node metastasis. We have previously published a method for studying CNAs in single DTCs by whole genome sequencing, where we compared two primary breast carcinomas to two corresponding DTCs. Copy number profiles from whole genome sequencing (WGS) from 40 DTCs were analyzed. The single cell whole genome amplified (WGA) DNA was used to generate WGS libraries, and the DTCs were subsequently sequenced on the Illumina HiSeq 2000. The WGS reads were trimmed for WGA adapters and aligned to GRCh37 human reference using Burrows-Wheeler Aligner (BWA). LogR values were calculated for genomic bins and corrected for% GC-bias and segmented using the piecewise constant fitting (PCF) algorithm (the penalty parameter, γ, was set to 25). Copy number was estimated per segment as 2logR × Ψ, where Ψ is the average ploidy. B allele frequency (BAF) was calculated for each known SNP position from dbSNP (dbSNP build 135) and somatic mutations read-outs generated. In this study we compared the mutation spectre and CNAs in six primary tumors, one with corresponding lymph node metastasis and single DTCs. In total, CN profile from 40 DTCs showed evidence of dissemination at both early and late stage of disease progression. At large, the copy number profile of the examined DTCs exhibited either a limited number of alterations, or a pattern similar to the primary tumor and lymph node metastasis suggesting continuous dissemination of single tumor cells throughout the tumor evolution. By demonstrating sub-clonality in the lymph node metastasis we provide novel insight into the metastatic process. Further, the correlation in aberration pattern between the lymph node metastasis and multiple DTCs, implies that cells found in the bone marrow may have originated from the lymph node metastasis. The DTCs exhibited common aberrations typically found in breast carcinomas, and several DTCs had deletion of 16q and17p, and gain of 1q, 8q. Certain DTCs exhibited CNAs not visible in the primary tumor or lymph node including gain of 9q, 14q, 19q and Xq, and loss of 2p, 6p, 8p, 18p and 19p. Two DTCs from time of diagnosis exhibited gain of the whole chromosome 5 that was not observed in the primary tumor or the lymph node. These results reveal the importance of assessing the sub-clonal genetic alterations in the primary tumor, as well as in the lymph node metastasis and DTCs, in order to evaluate patient treatment and prognosis. Citation Format: Elen Møller, Parveen Kumar, Silje Nord, David Wedge, Peter van Loo, April Peterson, Randi R. Mathiesen, Renathe Fjelldal, Masoud Z. Esteki, Jason A. Grundstad, Elin Borgen, Lars O. Baumbusch, Anne-Lise Børresen-Dale, Kevin P. White, Thierry Voet, Bjørn Naume, Vessela N. Kristensen. Tumor heterogeneity and dissemination in breast cancer: Deep sequencing of single disseminated cells from bone marrow compared to primary tumor and lymph node metastases. [abstract]. In: Proceedings of the 106th Annual Meeting of the American Association for Cancer Research; 2015 Apr 18-22; Philadelphia, PA. Philadelphia (PA): AACR; Cancer Res 2015;75(15 Suppl):Abstract nr LB-051. doi:10.1158/1538-7445.AM2015-LB-051

CNOGpro: detection and quantification of CNVs in prokaryotic whole-genome sequencing data

The explosion of whole-genome sequencing (WGS) as a tool in the mapping and understanding of genomes has been accompanied by an equally massive report of tools and pipelines for the analysis of DNA copy number variation (CNV). Most currently available tools are designed specifically for human genomes, with comparatively little literature devoted to CNVs in prokaryotic organisms. However, there are several idiosyncrasies in prokaryotic WGS data. This work proposes a step-by-step approach for detection and quantification of copy number variants specifically aimed at prokaryotes. After aligning WGS reads to a reference genome, we count the individual reads in a sliding window and normalize these counts for bias introduced by differences in GC content. We then investigate the coverage in two fundamentally different ways: (i) Employing a Hidden Markov Model and (ii) by repeated sampling with replacement (bootstrapping) on each individual gene. The latter bypasses the complex problem of breakpoint determination. To demonstrate our method, we apply it to real and simulated WGS data and benchmark it against two popular methods for CNV detection. The proposed methodology will in some cases represent a significant jump in accuracy from other current methods. CNOGpro is written entirely in the R programming language and is available from the CRAN repository (http://cran.r-project.org) under the GNU General Public License.

P090 : NEXT-GENERATION SEQUENCING-BASED HLA TYPING OF SALIVA AND BLOOD SAMPLES FROM THE SAME DONORS PRODUCES CONCORDANT TYPING RESULTS

Aim We present the results of next-generation sequencing (NGS) based HLA typing, using paired blood and saliva samples to demonstrate that typing results are independent of tissue source. We show that NGS reads from whole-genome sequencing (WGS) reads are also suitable for HLA typing, similarly to reads produced by well-targeted HLA kits albeit with lower precision. Methods Both saliva and blood samples were collected from two families. All samples were sequenced by both WGS and using a specific kit targeting the HLA-A, B, C and HLA-DRB1, -DQB1 genes. Reads obtained from the WGS kit were filtered for reads aligning to the IMGT/HLA database, and this subset was used as an input for the HLA typing algorithms. Two typing approaches were used, the first is based on read coverage statistics, the second method is building a consensus sequence that can be searched in the IMGT/HLA database. Results There was no significant difference in precision and accuracy between the results of blood and saliva samples. The choice of sequencing method had a much greater effect on precision. The different typing algorithms produced similar results, the most visible difference being the number of ambiguities reported. The two sequencing approaches gave nearly identical results, however the precision of typing obtained from WGS reads is lower due to a lower number of reads suitable for HLA typing. Nevertheless, four digit typing was possible in most cases. Targeting of HLA genes allowed 8 digit precision in some samples and ambiguity was lower for both typing algorithms. Targeted sequencing makes it possible to resolve phase ambiguities and detect novel alleles, however WGS was able to provide information about other genes present in IMGT/HLA (MICA, MICB TAP1, TAP2 etc.). Conclusion HLA typing results from blood and saliva samples are indistinguishable. The number and quality of reads obtained by WGS and targeted sequencing are significantly different. The quality control measures built-in the typing algorithm provide valuable threshold values for different statistics to reduce ambiguities. WGS with 30–40× average coverage is not suitable for high precision HLA typing – a targeted kit should be used to get 6 or 8 digit results – but the WGS approach can gather information about wider range of genes in the IMGT/HLA database.

A better sequence-read simulator program for metagenomics.

BackgroundThere are many programs available for generating simulated whole-genome shotgun sequence reads. The data generated by many of these programs follow predefined models, which limits their use to the authors' original intentions. For example, many models assume that read lengths follow a uniform or normal distribution. Other programs generate models from actual sequencing data, but are limited to reads from single-genome studies. To our knowledge, there are no programs that allow a user to generate simulated data following non-parametric read-length distributions and quality profiles based on empirically-derived information from metagenomics sequencing data.ResultsWe present BEAR (Better Emulation for Artificial Reads), a program that uses a machine-learning approach to generate reads with lengths and quality values that closely match empirically-derived distributions. BEAR can emulate reads from various sequencing platforms, including Illumina, 454, and Ion Torrent. BEAR requires minimal user input, as it automatically determines appropriate parameter settings from user-supplied data. BEAR also uses a unique method for deriving run-specific error rates, and extracts useful statistics from the metagenomic data itself, such as quality-error models. Many existing simulators are specific to a particular sequencing technology; however, BEAR is not restricted in this way. Because of its flexibility, BEAR is particularly useful for emulating the behaviour of technologies like Ion Torrent, for which no dedicated sequencing simulators are currently available. BEAR is also the first metagenomic sequencing simulator program that automates the process of generating abundances, which can be an arduous task.ConclusionsBEAR is useful for evaluating data processing tools in genomics. It has many advantages over existing comparable software, such as generating more realistic reads and being independent of sequencing technology, and has features particularly useful for metagenomics work.

Visualization and probability-based scoring of structural variants within repetitive sequences

Motivation: Repetitive sequences account for approximately half of the human genome. Accurately ascertaining sequences in these regions with next generation sequencers is challenging, and requires a different set of analytical techniques than for reads originating from unique sequences. Complicating the matter are repetitive regions subject to programmed rearrangements, as is the case with the antigen-binding domains in the Immunoglobulin (Ig) and T-cell receptor (TCR) loci.Results: We developed a probability-based score and visualization method to aid in distinguishing true structural variants from alignment artifacts. We demonstrate the usefulness of this method in its ability to separate real structural variants from false positives generated with existing upstream analysis tools. We validated our approach using both target-capture and whole-genome experiments. Capture sequencing reads were generated from primary lymphoid tumors, cancer cell lines and an EBV-transformed lymphoblast cell line over the Ig and TCR loci. Whole-genome sequencing reads were from a lymphoblastoid cell-line.Availability: We implement our method as an R package available at https://github.com/Eitan177/targetSeqView. Code to reproduce the figures and results are also available.Contact: ehalper2@jhmi.eduSupplementary information: Supplementary data are available at Bioinformatics online.

Whole Genome Sequencing Reveals a Widely Expressed Mutation (MYD88 L265P) with Oncogenic Activity in Waldenstrom's Macroglobulinemia

Abstract 300We performed whole genome sequencing (WGS) of lymphoplasmacytic cells from 30 Waldenstrom’s Macroglobulinemia (WM) patients, with paired normal tissue sequencing for 10 patients. Tumor and normal genomes were both sequenced to an average of 66X coverage of mapped individual reads. A recurring sequence variant at position 38182641 in chromosome 3p22.2 was identified which resulted in a single nucleotide change from T®C in the myeloid differentiation primary response (MYD88) gene, and a predicted non-synonymous change at amino acid position 265 from leucine to proline (L265P). This variant was the most common of a median of 3,419 (range 2,540–4,011) somatic variants identified by WGS in paired patients, and was present in tumor cells from all 10 paired patients, and 16 of 20 unpaired patients. For 22 of 26 patients, the MYD88 L265P variant was heterozygous, whereas in 4 patients an acquired UPD event at 3p22.2 resulted in homozygous presence of the variant in at least a subset of tumor cells. Sanger sequencing confirmed the presence of the MYD88 L265P variant in all 26 patient tumor samples revealed by WGS, as well as in one additional patient’s tumor sample that was not identified by the variant calling algorithms but for whom 12% of the WGS read level mappings showed the MYD88 L265P variant. In contrast, the MYD88 L265P variant was absent in normal paired tissues. Sanger sequencing therefore confirmed the somatic presence of the MYD88 L265P variant in tumor cells from 27 of 30 (90%) WM patients, and also identified this variant in BCWM.1 and MWCL-1 WM cells. In contrast, the MYD88 L265P variant was absent in CD138+ selected tumor cells from 8 of 8 multiple myeloma (MM) patients, and CD19+ cells from 12 of 12 healthy individuals, as well as in 7 of 8 patients with IgM monoclonal gammopathy of unknown significance (MGUS), in whom absence of the MYD88 L265P variant was further confirmed by TA cloning and sequencing of at least 100 clones. In the sole IgM MGUS patient in whom the MYD88 L265P mutation was detected, subsequent disease evolution occurred. Importantly, knock-down of MYD88 expression by lentiviral transduction led to loss of NF-κβ signaling and apoptosis of both BCWM.1 and MWCL-1 WM cells, with enhanced survival observed by complementary transduction with MYD88 L265P versus MYD88 wild type. The results of these studies therefore demonstrate a widely expressed somatic variant (MYD88 L265P) in malignant LPC of WM, whose presence confers oncogenic activity, and which may help distinguish WM disease from IgM MGUS or MM. Disclosures:No relevant conflicts of interest to declare.

THOR: targeted high-throughput ortholog reconstructor

Low-coverage genomes (LCGs) are becoming an increasingly important source of data for phylogenetic studies. However, assembly of these genomes is time consuming, difficult and lags behind sequence generation. THOR is a fast, stringent application for targeted reconstruction of sequence orthologs in unassembled LCGs. Using a 4x coverage set of mouse whole-genome sequence reads, THOR could partially or completely reconstruct 416/1000 human promoter ortholog regions in approximately 7.3 min/promoter. THOR's reconstruction rate improves markedly with both higher-coverage, and less divergent target species.

TranscriptSNPView: a genome-wide catalog of mouse coding variation

To the Editor: With the recent release of the genome-wide sequence for multiple inbred mouse strains1, and with resequencing data for a large number of additional strains entering the public domain (http://www.niehs.nih.gov/crg/cprc.htm), we are one step closer to being able to identify the underlying genetic variants responsible for the trait characteristics that define each strain. Here, we describe a genome-wide catalog of coding variation in the mouse genome that was developed using an extensive collection of mouse DNA sequence reads, including those recently released by Celera, data from dbSNP2 and resequencing data generated by Perlegen Sciences for the US National Institute of Environmental Health Sciences (NIEHS). To display these data, we developed a new software tool, TranscriptSNPView, which has been integrated into the Ensembl Genome Browser to take advantage of the evolving mouse genome assembly and the latest Ensembl3 and Vega gene predictions4. TranscriptSNPView can be accessed via the Ensembl Genome Browser (http://www.ensembl.org/Mus_musculus/transcriptsnpview). TranscriptSNPView displays coding SNP data from 48 mouse strains (Supplementary Table 1 online). Using the SNP calling algorithm ssahaSNP5, we computed over 50 million SNPs from the common laboratory Mus musculus strains A/J, DBA/2J, 129X1/SvJ and 129S1/SvImJ from whole-genome shotgun sequence reads generated by Celera, and from C3HeB/FeJ and NOD BAC-end sequence reads generated by the Wellcome Trust Sanger Institute. We also generated SNP calls from the Mus musculus molossinus strain MSM/Ms using sequence reads generated by RIKEN6 (Supplementary Table 1). Collectively, these SNP calls have been designated ‘Sanger SNPs’. The 25 million DNA sequence reads used to generate the Sanger SNP collection represent 7.32-fold coverage of the NCBI mouse build 35 genome assembly and are available via the Ensembl trace repository (http://trace.ensembl.org). The Sanger SNP calls were distilled to 6.87 million nonredundant genome-wide SNP features and were combined with an additional 6.4 million dbSNP entries (version 126), providing data for an additional 41 mouse strains. By merging these data sets and mapping them against the Ensembl 38.35 mouse gene build, we collated 726,462 coding SNP variants across all strains and computed their amino acid consequences to identify 249,996 nonsynonymous coding changes and 2,667 stop codons. Coding SNP figures for each strain are provided in Supplementary Table 1. We also identified instances where stop codons had been lost, and we predicted mutations in introns, invariant intronic splice sites and in untranslated and regulatory regions. These predictions, which can be used as a basis for identifying functional SNP variants, are displayed in TranscriptSNPView. A detailed description of all of the features of TranscriptSNPView is provided in the Supplementary Note online. A data collection of this quality and depth is unprecedented and will provide the means to obtain a high-resolution picture of coding variation in the mouse genome. TrancriptSNPView represents a powerful new tool for functional analysis of the mouse genome and will become a central repository for mouse coding variation data.