A robust benchmark for germline structural variant detection

Zook, Justin M.; Nf, Hansen; Nd, Olson; Lm, Chapman; Jc, Mullikin; Xiao, Chunlin; Sherry, Stephen T.; Koren, Sergey; Am, Phillippy; Boutros, Paul C.; Sme, Sahraeian; Huang, Biao; Rouette, Alexandre; Alexander, Noah; Ce, Mason; Hajirasouliha, Iman; Ricketts, Camir; J, Lee; Tearle, Rick; It, Fiddes; Am, Barrio; Wala, Jeremiah A.; Carroll, Andrew; Ghaffari, Noushin; Rodriguez, Oscar L.; Bashir, Ali Kashif; Jackman, Shaun D.; Jj, Farrell; Am, Wenger; Alkan, Can; Söylev, Arda; Mc, Schatz; Garg, Shilpa; Church, George M.; Marschall, Tobias; Chen, K; Fan, Xing; Ac, English; Ja, Rosenfeld; Zhou, Weichen; Mills, Ryan E.; Jm, Sage; Davis, W. Rhett; Kaiser,; Js, Oliver; Ap, Catalano; Mj, Chaisson; Spies, Noah; Sedlazeck, Fritz J.; Salit, Marc

doi:10.1101/664623

Cited by 56 publications

(64 citation statements)

References 81 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For SNVs, we benchmarked the calls from three strategies using the gold standard of NA12878 (ftp://ftp-trace.ncbi.nlm.nih.gov/giab/ftp/release/NA12878_HG001/latest/GRCh38/) and NA24385 (ftp://ftp-trace.ncbi.nlm.nih.gov/giab/ftp/release/AshkenazimTrio/HG002_NA24385_son/latest/GRCh38/). For SVs, we compared three linked-read sets (R 9 , R 10 , R 11 ) from HG002 with the Tier 1 SV benchmark from Genome in a Bottle [36] and used VaPoR [37] to validate our SV calls based on PacBio CCS reads from NA24385 (Highly-accurate long-read sequencing improves variant detection and assembly of a human genome). We compared SNV and SV calls among the different approaches using vcfeval [38] and truvari [36], respectively.…”

Section: Data Descriptionmentioning

confidence: 99%

Assessment of human diploid genome assembly with 10x Linked-Reads data

Zhang

Zhou

Weng

et al. 2019

Preprint

View full text Add to dashboard Cite

14Background: Producing cost-effective haplotype-resolved personal genomes remains 15 challenging. 10x Linked-Read sequencing, with its high base quality and long-range information, 16 has been demonstrated to facilitate de novo assembly of human genomes and variant detection. 17In this study, we investigate in depth how the parameter space of 10x library preparation and 18 sequencing affects assembly quality, on the basis of both simulated and real libraries. 19Findings: We prepared and sequenced eight 10x libraries with a diverse set of parameters from 20 standard cell lines NA12878 and NA24385 and performed whole genome assembly on the data. 21We also developed the simulator LRTK-SIM to follow the workflow of 10x data generation and 22 produce realistic simulated Linked-Read data sets. We found that assembly quality could be 23 improved by increasing the total sequencing coverage (C) and keeping physical coverage of 24 DNA fragments (C F ) or read coverage per fragment (C R ) within broad ranges. The optimal 25 physical coverage was between 332X and 823X and assembly quality worsened if it increased 26 to greater than 1,000X for a given C. Long DNA fragments could significantly extend phase 27 blocks, but decreased contig contiguity. The optimal length-weighted fragment length (Wߤ ி ) 28 was around 50 -150kb. When broadly optimal parameters were used for library preparation 29 and sequencing, ca. 80% of the genome was assembled in a diploid state. 30Conclusion: The Linked-Read libraries we generated and the parameter space we identified 31 provide theoretical considerations and practical guidelines for personal genome assemblies 32 based on 10x Linked-Read sequencing. 33 Keywords: 10x Linked-Read sequencing, de novo assembly, diploid human genome, library 34 preparation 35 3 Data description 36 Introduction 37 The human genome holds the key for understanding the genetic basis of human evolution, 38 hereditary illnesses and many phenotypes. Whole-genome reconstruction and variant discovery, 39 accomplished by analysis of data from whole-genome sequencing experiments, are 40 foundational for the study of human genomic variation and analysis of genotype-phenotype 41 relationships. Over the past decades, cost-effective whole-genome sequencing has been 42 revolutionized by short-fragment approaches, the most widespread of which have been the 43 consistently improving generations of the original Solexa technology [1, 2], now referred to as 44 Illumina sequencing. Illumina's strengths and weaknesses are inherent in the sample 45 preparation and sequencing chemistry. Illumina generates short paired reads (2x150 base pairs 46 for the highest-throughput platforms) from short fragments (usually 400-500 base pairs) [3]. 47 Because many clonally amplified molecules generate a robust signal during the sequencing 48 reaction, Illumina's average per-base error rates are very low. 50The lack of long-range contiguity between end-sequenced short fragments limits their 51 application for reconstructing personal genomes. Long-rang...

show abstract

Section: Data Descriptionmentioning

confidence: 99%

Assessment of human diploid genome assembly with 10x Linked-Reads data

Zhang

Zhou

Weng

et al. 2019

Preprint

View full text Add to dashboard Cite

show abstract

“…10 Because of the unique value gained by optical mapping of ultra-long DNA reads, it has been used in essentially all modern reference genome assemblies (human GRCh, 11; 12 mouse, 13 goat, 14 maize, 15 as well as benchmark structural variation papers. 16-18…”

Section: Introductionmentioning

confidence: 99%

Next generation cytogenetics: comprehensive assessment of 48 leukemia genomes by genome imaging

Neveling

Mantere

Vermeulen

et al. 2020

Preprint

View full text Add to dashboard Cite

Somatic structural variants are important for cancer development and progression. In a diagnostic set-up, especially for hematological malignancies, the comprehensive analysis of all cytogenetic aberrations in a given sample still requires a combination of techniques, such as karyotyping, fluorescence in situ hybridization and CNV-microarrays. We hypothesize that the combination of these classical approaches could be replaced by high-resolution genome imaging.Bone marrow aspirates or blood samples derived from 48 patients with leukemia, who received a clinical diagnoses of different types of hematological malignancies, were processed for genome imaging with the Bionano Genomics Saphyr system. In all cases cytogenetic abnormalities had previously been identified using standard of care workflows. Based on these diagnostic results, the samples were divided into two categories: simple cases (<5 aberrations, n=37) and complex cases (≥5 aberrations or an unspecified marker chromosome, n=11). By imaging the labelled ultra-long gDNA molecules (average N50 >250kb), we generated on average ∼280-fold mapped genome coverage per sample. Chromosomal aberrations were called by Bionano Genomics Rare variant pipeline (RVP) specialized for the detections of somatic variants.Per sample, on average a total of 1,454 high confidence SVs were called, and on average 44 (range: 14-130) of those were rare i.e. not present in the population control database. Importantly, for the simple cases, all clinically reported aberrations with variant allele frequencies higher than 10% were detected by genome imaging. This held true for deletions, insertions, inversions, aneuploidies and translocations. The results for the complex cases were also largely concordant between the standard of care workflow and optical mapping, and in several cases, optical mapping revealed higher complexity than previously known. SV and CNV calls detected by optical mapping were more complete than any other previous single test and likely delivered the most accurate and complete underlying genomic architecture. Even complex chromothripsis structures were resolved. Finally, optical mapping also identified multiple novel events, including balanced translocations that lead to potential novel fusion-genes, opening the potential to discover new prognostic and diagnostic biomarkers.The full concordance with diagnostic standard assays for simple cases and the overall great concordance with (previously likely incompletely understood) complex cases demonstrates the potential to replace classical cytogenetic tests with genome imaging. In addition, this holds the potential to rapidly map new fusion genes and identify novel SVs and CNVs as novel potential leukemia drivers.

show abstract

“…We describe a machine learning model, BioGraph QUALclassifier, that uses coverage signatures to assign quality scores to discovered alleles to increase specificity and assist prioritization of variants. We benchmark five technical replicates of an individual (HG002) against the Genome in a Bottle Tier 1 SV set [ 21 ] alongside other SV detection pipelines to illustrate relative performance. Finally, we merge discovered calls and confirm the presence of alleles with BioGraph Coverage to assess the per-replicate increase in recall over discovery.…”

Section: Introductionmentioning

confidence: 99%

Leveraging a WGS compression and indexing format with dynamic graph references to call structural variants

English¹,

McCarthy²,

Flickenger³

et al. 2020

Preprint

Self Cite

View full text Add to dashboard Cite

We introduce a novel structural variant calling pipeline (BioGraph) that leverages a read compression and indexing format to discover alleles, assess their supporting coverage, and assign useful quality scores. To evaluate BioGraph's performance, we run five sequencing replicates of the individual HG002 through the BioGraph variant calling pipeline, as well as two other short-read sv-calling pipelines. BioGraph detects the GIAB benchmark SVs at a peak sensitivity of ≈59% compared to ≈42% sensitivity from the other pipelines. The overall precision of BioGraph is lower than other pipelines (≈81% and ≈90%, respectively), however, adjusting for quality score, BioGraph calls were sensitive to a greater number of SV calls given the same false discovery rate compared to the other pipelines. Cumulatively ≈77% of GIAB benchmark SVs are discovered by BioGraph in at least one replicate. After merging discovered calls and running BioGraph Coverage to create a squared-off project-level VCF, we find ≈90% percent of discovered true positive alleles have at least 5x coverage in all replicates, thus increasing per-replicate recall of alleles having at least 1x coverage to ≈76.9%.

show abstract

A robust benchmark for germline structural variant detection

Cited by 56 publications

References 81 publications

Assessment of human diploid genome assembly with 10x Linked-Reads data

Assessment of human diploid genome assembly with 10x Linked-Reads data

Next generation cytogenetics: comprehensive assessment of 48 leukemia genomes by genome imaging

Leveraging a WGS compression and indexing format with dynamic graph references to call structural variants

Contact Info

Product

Resources

About