Recent improvements in sequencing chemistry and instrument performance combine to create a new PacBio data type, Single Molecule High-Fidelity reads (HiFi reads). Increased read length and improvement in library construction enables average read lengths of 10-20 kb with average sequence identity greater than 99% from raw single molecule reads. The resulting reads have the accuracy comparable to short read NGS but with 50-100 times longer read length. Here we benchmark the performance of this data type by sequencing and genotyping the Genome in a Bottle (GIAB) HG0002 human reference sample from the National Institute of Standards and Technology (NIST). We further demonstrate the general utility of HiFi reads by analyzing multiple clones of Cabernet Sauvignon. Three different clones were sequenced and de novo assembled with the CANU assembly algorithm, generating draft assemblies of very high contiguity equal to or better than earlier assembly efforts using PacBio long reads. Using the Cabernet Sauvignon Clone 8 assembly as a reference, we mapped the HiFi reads generated from Clone 6 and Clone 47 to identify single nucleotide polymorphisms (SNPs) and structural variants (SVs) that are specific to each of the three samples.
Structural variant detection with long read sequencing reveals driver and passenger mutations in a melanoma cell line
Past large scale cancer genome sequencing efforts, including The Cancer Genome Atlas and the International Cancer Genome Consortium, have utilized short-read sequencing, which is well-suited for detecting single nucleotide variants (SNVs) but far less reliable for detecting variants larger than 20 base pairs, including insertions, deletions, duplications, inversions and translocations. Recent same-sample comparisons of short- and long-read human reference genome data have revealed that short-read resequencing typically uncovers only ~4,000 structural variants (SVs, =50 bp) per genome and is biased towards deletions, whereas sequencing with PacBio long-reads consistently finds ~20,000 SVs, evenly balanced between insertions and deletions. This discovery has important implications for cancer research, as it is clear that SVs are both common and biologically important in many cancer subtypes, including colorectal, breast and ovarian cancer. Without confident and comprehensive detection of structural variants, it is unlikely we have a sufficiently complete picture of all the genomic changes that impact cancer development, disease progression, treatment response, drug resistance, and relapse. To begin to address this unmet need, we have sequenced the COLO829 tumor and matched normal lymphoblastoid cell lines to 49- and 51-fold coverage, respectively, with PacBio SMRT Sequencing, with the goal of developing a high-confidence structural variant call set that can be used to empirically evaluate cost-effective experimental designs for larger scale studies and develop structural variation calling software suitable for cancer genomics. Structural variant calling revealed over 21,000 deletions and 19,500 insertions larger than 20 bp, nearly four times the number of events detected with short-read sequencing. The vast majority of events are shared between the tumor and normal, with about 100 putative somatic deletions and 400 insertions, primarily in microsatellites. A further 40 rearrangements were detected, nearly exclusively in the tumor. One rearrangement is shared between the tumor and normal, t(5;X) which disrupts the mismatch repeat gene MSH3, and is likely a driver mutation. Generating high-confidence call sets that cover the entire size-spectrum of somatic variants from a range of cancer model systems is the first step in determining what will be the best approach for addressing an ongoing blind spot in our current understanding of cancer genomes. Here the application of PacBio sequencing to a melanoma cancer cell line revealed thousands of previously overlooked variants, including a mutation likely involved in tumorogenesis.
Genomics studies have shown that the insertions, deletions, duplications, translocations, inversions, and tandem repeat expansions in the structural variant (SV) size range (>50 bp) contribute to the evolution of traits and often have significant associations with agronomically important phenotypes. However, most SVs are too small to detect with array comparative genomic hybridization and too large to reliably discover with short-read DNA sequencing. While de novo assembly is the most comprehensive way to identify variants in a genome, recent studies in human genomes show that PacBio SMRT Sequencing sensitively detects structural variants at low coverage. Here we present SV characterization in the major crop species Oryza sativa subsp. indica (rice) with low-fold coverage of long reads. In addition, we provide recommendations for sequencing and analysis for the application of this workflow to other important agricultural species.
To comprehensively detect large variants in human genomes, we have extended pbsv – a structural variant caller for long reads – to call copy-number variants (CNVs) from read-clipping and read-depth signatures. In human germline benchmark samples, we detect more than 300 CNVs spanning around 10 Mb, and we call hundreds of additional events in re-arranged cancer samples. Long-read sequencing of diverse humans has revealed more than 20,000 insertion, deletion, and inversion structural variants spanning more than 12 Mb in a typical human genome. Most of these variants are too large to detect with short reads and too small for array comparative genome hybridization (aCGH). While the standard approaches to calling structural variants with long reads thrive in the 50 bp to 10 kb size range, they tend to miss exactly the large (>50 kb) copy-number variants that are called more readily with aCGH and short reads. Standard algorithms rely on reference-based mapping of reads that fully span a variant or on de novo assembly; and copy-number variants are often too large to be spanned by a single read and frequently involve segmentally duplicated sequence that is not yet included in most de novo assemblies.
Detection and phasing of small variants in Genome in a Bottle samples with highly accurate long reads
Introduction: Long-read PacBio SMRT Sequencing has been applied successfully to assemble genomes and detect structural variants. However, due to high raw read error rates of 10-15%, it has remained difficult to call small variants from long reads. Recent improvements in library preparation, sequencing chemistry, and instrument yield have increased length, accuracy, and throughput of PacBio Circular Consensus (CCS) reads, resulting in 10-20 kb “HiFi” reads with mean read quality above 99%. Materials and Methods: We sequenced 11 kb size-selected libraries from the Genome in a Bottle (GIAB) human reference samples HG001, HG002, and HG005 to approximately 30-fold coverage on the Sequel II System with six SMRT Cells 8M each. The CCS algorithm was used to generate highly accurate (average 99.8%) reads of mean length 10-11 kb, which were then mapped to the hs37d5 reference with pbmm2. We detected small variants using Google DeepVariant and compared these variant calls to GIAB benchmarks. Small variants were then phased with WhatsHap. Results: With these long, highly accurate CCS reads, DeepVariant achieves high SNP and Indel accuracy against the GIAB benchmark truth set for all three reference samples. Using WhatsHap, small variants were phased into haplotype blocks with N50 from 82 to 146 kb. The improved mappability of long reads allows detection of variants in many medically relevant genes such as CYP2D6and PMS2that have proven ‘difficult-to-map’ with short reads. We show that small variant precision and recall remain high down to 15-fold coverage. Conclusions: These highly accurate long reads combine the mappability of noisy long reads with the accuracy and small variant detection utility of short reads, which will allow the detection and phasing of variants in regions that have proven recalcitrant to short read sequencing and variant detection.
Introduction: Long-read sequencing has been applied successfully to assemble genomes and detect structural variants. However, due to high raw-read error rates (10-15%), it has remained difficult to call small variants from long reads. Recent improvements in library preparation and sequencing chemistry have increased length, accuracy, and throughput of PacBio circular consensus sequencing (CCS) reads, resulting in 15-20kb reads with average read quality above 99%. Materials and Methods: We sequenced a library from human reference sample HG002 to 18-fold coverage on the PacBio Sequel II with two SMRT Cells 8M. The CCS algorithm was used to generate highly accurate (average 99.9%) 12.9kb reads, which were mapped to the hg19 reference with pbmm2. We detected small variants using Google DeepVariant with a model trained for CCS and phased the variants using WhatsHap. Structural variants were detected with pbsv. Variant calls were evaluated against Genome in a Bottle (GIAB) benchmarks. Results: With these reads, DeepVariant achieves SNP and Indel F1 scores of 99.70% and 96.59% against the GIAB truth set, and pbsv achieves 97.72% recall on structural variants longer than 50bp. Using WhatsHap, small variants were phased into haplotype blocks with 145kb N50. The improved mappability of long reads allows us to align to and detect variants in medically relevant genes such as CYP2D6 and PMS2 that have proven “difficult-to-map” with short reads. Conclusions: These highly accurate long reads combine the mappability and ability to detect structural variants of long reads with the accuracy and ability to detect small variants of short reads.
New technologies and analysis methods are enabling genomic structural variants (SVs) to be detected with ever-increasing accuracy, resolution, and comprehensiveness. Translating these methods to routine research and clinical practice requires robust benchmark sets. We developed the first benchmark set for identification of both false negative and false positive germline SVs, which complements recent efforts emphasizing increasingly comprehensive characterization of SVs. To create this benchmark for a broadly consented son in a Personal Genome Project trio with broadly available cells and DNA, the Genome in a Bottle (GIAB) Consortium integrated 19 sequence-resolved variant calling methods, both alignment- and de novo assembly-based, from short-, linked-, and long-read sequencing, as well as optical and electronic mapping. The final benchmark set contains 12745 isolated, sequence-resolved insertion and deletion calls =50 base pairs (bp) discovered by at least 2 technologies or 5 callsets, genotyped as heterozygous or homozygous variants by long reads. The Tier 1 benchmark regions, for which any extra calls are putative false positives, cover 2.66 Gbp and 9641 SVs supported by at least one diploid assembly. Support for SVs was assessed using svviz with short-, linked-, and long-read sequence data. In general, there was strong support from multiple technologies for the benchmark SVs, with 90 % of the Tier 1 SVs having support in reads from more than one technology. The Mendelian genotype error rate was 0.3 %, and genotype concordance with manual curation was >98.7 %. We demonstrate the utility of the benchmark set by showing it reliably identifies both false negatives and false positives in high-quality SV callsets from short-, linked-, and long-read sequencing and optical mapping.
Benchmark small variant calls are required for developing, optimizing and assessing the performance of sequencing and bioinformatics methods. Here, as part of the Genome in a Bottle (GIAB) Consortium, we apply a reproducible, cloud-based pipeline to integrate multiple short- and linked-read sequencing datasets and provide benchmark calls for human genomes. We generate benchmark calls for one previously analyzed GIAB sample, as well as six genomes from the Personal Genome Project. These new genomes have broad, open consent, making this a ‘first of its kind’ resource that is available to the community for multiple downstream applications. We produce 17% more benchmark single nucleotide variations, 176% more indels and 12% larger benchmark regions than previously published GIAB benchmarks. We demonstrate that this benchmark reliably identifies errors in existing callsets and highlight challenges in interpreting performance metrics when using benchmarks that are not perfect or comprehensive. Finally, we identify strengths and weaknesses of callsets by stratifying performance according to variant type and genome context.
Single-molecule long-read sequencing datasets were generated for a son-father-mother trio of Han Chinese descent that is part of the Genome in a Bottle (GIAB) consortium portfolio. The dataset was generated using the Pacific Biosciences Sequel System. The son and each parent were sequenced to an average coverage of 60 and 30, respectively, with N50 subread lengths between 16 and 18?kb. Raw reads and reads aligned to both the GRCh37 and GRCh38 are available at the NCBI GIAB ftp site (ftp://ftp-trace.ncbi.nlm.nih.gov/giab/ftp/data/ChineseTrio/). The GRCh38 aligned read data are archived in NCBI SRA (SRX4739017, SRX4739121, and SRX4739122). This dataset is available for anyone to develop and evaluate long-read bioinformatics methods.
The number of human genomes being genotyped or sequenced increases exponentially and efficient haplotype estimation methods able to handle this amount of data are now required. Here, we present a new method, SHAPEIT4, which substantially improves upon other methods to process large genotype and high coverage sequencing datasets. It notably exhibits sub-linear scaling with sample size, provides highly accurate haplotypes and allows integrating external phasing information such as large reference panels of haplotypes, collections of pre-phased variants and long sequencing reads. We provide SHAPET4 in an open source format on https://odelaneau.github.io/shapeit4/ and demonstrate its performance in terms of accuracy and running times on two gold standard datasets: the UK Biobank data and the Genome In A Bottle.
Application of assembly methods for personal genome analysis from next generation sequencing data has been limited by the requirement for an expensive supercomputer hardware or long computation times when using ordinary resources. We describe CompStor Novos, achieving supercomputer-class performance in de novo assembly computation time on standard server hardware, based on a tiered-memory algorithm. Run on commercial off-the-shelf servers, Novos assembly is more precise and 10-20 times faster than that of existing assembly algorithms. Furthermore, we integrated Novos into a variant calling pipeline and demonstrate that both compute times and precision of calling point variants and indels compare well with standard alignment-based pipelines. Additionally, assembly eliminates bias in the estimation of allele frequency for indels and naturally enables discovery of breakpoints for structural variants with base pair resolution. Thus, Novos bridges the gap between alignment-based and assembly-based genome analyses. Extension and adaption of its underlying algorithm will help quickly and fully harvest information in sequencing reads for personal genome reconstruction.
Automated structural variant verification in human genomesw using single-molecule electronic DNA mapping.
The importance of structural variation in human disease and the difficulty of detecting structural variants larger than 50 base pairs has led to the development of several long-read sequencing technologies and optical mapping platforms. Frequently, multiple technologies and ad hoc methods are required to obtain a consensus regarding the location, size and nature of a structural variant, with no approach able to reliably bridge the gap of variant sizes between the domain of short-read approaches and the largest rearrangements observed with optical mapping. To address this unmet need, we have developed a new software package, SV-VerifyTM, which utilizes data collected with the Nabsys High Definition Mapping (HD-MappingTM) system, to perform hypothesis-based verification of putative deletions. We demonstrate that whole genome maps, constructed from electronic detection of tagged DNA, hundreds of kilobases in length, can be used effectively to facilitate calling of structural variants ranging in size from 300 base pairs to hundreds of kilobase pairs. SV-Verify implements hypothesis-based verification of putative structural variants using a set of support vector machines and is capable of concurrently testing several thousand independent hypotheses. We describe support vector machine training, utilizing a well-characterized human genome, and application of the resulting classifiers to another human genome, demonstrating high sensitivity and specificity for deletions >= 300 base pairs.
Read-based phasing deduces the haplotypes of an individual from sequencing reads that cover multiple variants, while genetic phasing takes only genotypes as input and applies the rules of Mendelian inheritance to infer haplotypes within a pedigree of individuals. Combining both into an approach that uses these two independent sources of information-reads and pedigree-has the potential to deliver results better than each individually.We provide a theoretical framework combining read-based phasing with genetic haplotyping, and describe a fixed-parameter algorithm and its implementation for finding an optimal solution. We show that leveraging reads of related individuals jointly in this way yields more phased variants and at a higher accuracy than when phased separately, both in simulated and real data. Coverages as low as 2× for each member of a trio yield haplotypes that are as accurate as when analyzed separately at 15× coverage per individual.https://email@example.com.© The Author 2016. Published by Oxford University Press.
Complex chromosomal rearrangements are structural genomic alterations involving multiple instances of deletions, duplications, inversions, or translocations that co-occur either on the same chromosome or represent different overlapping events on homologous chromosomes. We present SVelter, an algorithm that identifies regions of the genome suspected to harbor a complex event and then resolves the structure by iteratively rearranging the local genome structure, in a randomized fashion, with each structure scored against characteristics of the observed sequencing data. SVelter is able to accurately reconstruct complex chromosomal rearrangements when compared to well-characterized genomes that have been deeply sequenced with both short and long reads.
We report on the sequencing of 10,545 human genomes at 30×-40× coverage with an emphasis on quality metrics and novel variant and sequence discovery. We find that 84% of an individual human genome can be sequenced confidently. This high-confidence region includes 91.5% of exon sequence and 95.2% of known pathogenic variant positions. We present the distribution of over 150 million single-nucleotide variants in the coding and noncoding genome. Each newly sequenced genome contributes an average of 8,579 novel variants. In addition, each genome carries on average 0.7 Mb of sequence that is not found in the main build of the hg38 reference genome. The density of this catalog of variation allowed us to construct high-resolution profiles that define genomic sites that are highly intolerant of genetic variation. These results indicate that the data generated by deep genome sequencing is of the quality necessary for clinical use.