Structural Variants (SVs), which include deletions, insertions, duplications, inversions and chromosomal rearrangements, have been shown to effect organism phenotypes, including changing gene expression, increasing disease risk, and playing an important role in cancer development. Still it remains challenging to detect all types of SVs from high throughput sequencing data and it is even harder to detect more complex SVs such as a duplication nested within an inversion. To overcome these challenges we developed algorithms for SV analysis using longer third generation sequencing reads. The increased read lengths allow us to span more complex SVs and accurately assess SVs in repetitive regions, two of the major limitations when using short Illumina data. Our enhanced open-source analysis method Sniffles accurately detects structural variants based on split read mapping and assessment of the alignments. Sniffles uses a self-balancing interval tree in combination with a plane sweep algorithm to manage and assess the identified SVs. Central to its high accuracy is its advanced scoring model that can distinguish erroneous alignments from true breakpoints flanking SVs. In experiments with simulated and real genomes (e.g human breast cancer), we find that Sniffles outperforms all other SV analysis approaches in both the sensitivity of finding events as well as the specificity of those events. Sniffles is available at: https://github.com/fritzsedlazeck/Sniffles
Comprehensive genome and transcriptome structural analysis of a breast cancer cell line using PacBio long read sequencing
Genomic instability is one of the hallmarks of cancer, leading to widespread copy number variations, chromosomal fusions, and other structural variations. The breast cancer cell line SK-BR-3 is an important model for HER2+ breast cancers, which are among the most aggressive forms of the disease and affect one in five cases. Through short read sequencing, copy number arrays, and other technologies, the genome of SK-BR-3 is known to be highly rearranged with many copy number variations, including an approximately twenty-fold amplification of the HER2 oncogene. However, these technologies cannot precisely characterize the nature and context of the identified genomic events and other important mutations may be missed altogether because of repeats, multi-mapping reads, and the failure to reliably anchor alignments to both sides of a variation. To address these challenges, we have sequenced SK-BR-3 using PacBio long read technology. Using the new P6-C4 chemistry, we generated more than 70X coverage of the genome with average read lengths of 9-13kb (max: 71kb). Using Lumpy for split-read alignment analysis, as well as our novel assembly-based algorithms for finding complex variants, we have developed a detailed map of structural variations in this cell line. Taking advantage of the newly identified breakpoints and combining these with copy number assignments, we have developed an algorithm to reconstruct the mutational history of this cancer genome. From this we have discovered a complex series of nested duplications and translocations between chr17 and chr8, two of the most frequent translocation partners in primary breast cancers, resulting in amplification of HER2. We have also carried out full-length transcriptome sequencing using PacBio’s Iso-Seq technology, which has revealed a number of previously unrecognized gene fusions and isoforms. Combining long-read genome and transcriptome sequencing technologies enables an in-depth analysis of how changes in the genome affect the transcriptome, including how gene fusions are created across multiple chromosomes. This analysis has established the most complete cancer reference genome available to date, and is already opening the door to applying long-read sequencing to patient samples with complex genome structures.
Despite amazing progress over the past quarter century in the technology to detect genetic variants, intermediate-sized structural variants (50 bp to 50 kb) have remained difficult to identify. Such variants are too small to detect with array comparative genomic hybridization, but too large to reliably discover with short-read DNA sequencing. Recent de novo assemblies of human genomes have demonstrated the power of PacBio Single Molecule, Real-Time (SMRT) Sequencing to fill this technology gap and sensitively identify structural variants in the human genome. While de novo assembly is the ideal method to identify variants in a genome, it requires high depth of coverage. A structural variant discovery approach that utilizes lower coverage would facilitate evaluation of large patient and population cohorts. Here we introduce such an approach and apply it to 10-fold coverage of several human genomes generated on the PacBio Sequel System. To identify structural variants in low-fold coverage whole genome sequencing data, we apply a reference-based, re-sequencing workflow. First, reads are mapped to the human reference genome with a local aligner. The local alignments often end at structural variant loci. To connect co-linear local alignments across structural variants, we apply a novel algorithm that merges alignments into “chains” and refines the alignment edges. Then, the chained alignments are scanned for windows with an excess of insertions or deletions to identify candidate structural variant loci. Finally, the read support at each putative variant locus is evaluated to produce a variant call. Single nucleotide information is incorporated to phase and evaluate the zygosity of each structural variant. In 10-fold coverage human genome sequence, we identify the vast majority of the structural variants found by de novo assembly, thus demonstrating the power of low-fold coverage SMRT Sequencing to affordably and effectively detect structural variants.
Though a role for structural variants in human disease has long been recognized, it has remained difficult to identify intermediate-sized variants (50 bp to 5 kb), which are too small to detect with array comparative genomic hybridization, but too large to reliably discover with short-read DNA sequencing. Recent studies have demonstrated that PacBio Single Molecule, Real-Time (SMRT) sequencing fills this technology gap. SMRT sequencing detects tens of thousands of structural variants in the human genome, approximately five times the sensitivity of short-read DNA sequencing.
Structural variants (genomic differences =50 base pairs) contribute to the evolution of organisms traits and human disease. Most structural variants (SVs) are too small to detect with array comparative genomic hybridization but too large to reliably discover with short-read DNA sequencing. Recent studies in human genomes show that PacBio SMRT Sequencing sensitively detects structural variants.
Most of the base pairs that differ between two human genomes are in intermediate-sized structural variants (50 bp to 5 kb), which are too small to detect with array comparative genomic hybridization or optical mapping but too large to reliably discover with short-read DNA sequencing. Long-read sequencing with PacBio Single Molecule, Real-Time (SMRT) Sequencing platforms fills this technology gap. PacBio SMRT Sequencing detects tens of thousands of structural variants in a human genome with approximately five times the sensitivity of short-read DNA sequencing. Effective application of PacBio SMRT Sequencing to detect structural variants requires quality bioinformatics tools that account for the characteristics of PacBio reads. To provide such a solution, we developed pbsv, a structural variant caller for PacBio reads that works as a chain of simple stages: 1) map reads to the reference genome, 2) identify reads with signatures of structural variation, 3) cluster nearby reads with similar signatures, 4) summarize each cluster into a consensus variant, and 5) filter for variants with sufficient read support. To evaluate the baseline performance of pbsv, we generated high coverage of a diploid human genome on the PacBio Sequel System, established a target set of structural variants, and then titrated to lower coverage levels. The false discovery rate for pbsv is low at all coverage levels. Sensitivity is high even at modest coverage: above 85% at 10-fold coverage and above 95% at 20-fold coverage. To assess the potential for PacBio SMRT Sequencing to identify pathogenic variants, we evaluated an individual with clinical symptoms suggestive of Carney complex for whom short-read whole genome sequencing was uninformative. The individual was sequenced to 9-fold coverage on the PacBio Sequel System, and structural variants were called with pbsv. Filtering for rare, genic structural variants left six candidates, including a heterozygous 2,184 bp deletion that removes the first coding exon of PRKAR1A. Null mutations in PRKAR1Acause autosomal dominant Carney complex, type 1. The variant was determined to be de novo, and it was classified as likely pathogenic based on ACMG standards and guidelines for variant interpretation. These case studies demonstrate the ability of pbsv to detect structural variants in low-coverage PacBio SMRT Sequencing and suggest the importance of considering structural variants in any study of human genetic variation.
Structural variants (genomic differences =50 base pairs) contribute to the evolution of traits and disease. Most structural variants (SVs) are too small to detect with array comparative genomic hybridization and too large to reliably discover with short-read DNA sequencing.
Allelic specificity of immunoglobulin heavy chain ([email protected]) translocation in B-cell acute lymphoblastic leukemia (B-ALL) unveiled by long-read sequencing
Oncogenic fusion of IGH-DUX4 has recently been reported as a hallmark that defines a B-ALL subtype present in up to 7% of adolescents and young adults B-ALL. The translocation of DUX4 into IGH results in aberrant activation of DUX4 by hijacking the intronic IGH enhancer (Eµ). How IGH-DUX4 translocation interplays with IGH allelic exclusion was never been explored. We investigated this in Nalm6 B-ALL cell line, using long-read (PacBio Iso-Seq method and 10X Chromium WGS), short-read (Illumina total stranded RNA and WGS), epigenome (H3K27ac ChIP-seq, ATAC-seq) and 3-D genome (Hi-C, H3K27ac HiChIP, Capture-C).
Structural variants (SVs) – genomic differences =50 base pairs – are few by count compared to single nucleotide variants (SNVs) and indels but include most of the base pairs that differ between two humans.
Although the accuracy of the human reference genome is critical for basic and clinical research, structural variants (SVs) have been difficult to assess because data capable of resolving them have been limited. To address potential bias, we sequenced a diversity panel of nine human genomes to high depth using long-read, single-molecule, real-time sequencing data. Systematically identifying and merging SVs =50 bp in length for these nine and one public genome yielded 83,909 sequence-resolved insertions, deletions, and inversions. Among these, 2,839 (2.0 Mbp) are shared among all discovery genomes with an additional 13,349 (6.9 Mbp) present in the majority of humans, indicating minor alleles or errors in the reference, which is partially explained by an enrichment for GC-content and repetitive DNA. Genotyping 83% of these in 290 additional genomes confirms that at least one allele of the most common SVs in unique euchromatin are now sequence-resolved. We observe a 9-fold increase within 5 Mbp of chromosome telomeric ends and correlation with de novo single-nucleotide variant mutations showing that such variation is nonrandomly distributed defining potential hotspots of mutation. We identify SVs affecting coding and noncoding regulatory loci improving annotation and interpretation of functional variation. To illustrate the utility of sequence-resolved SVs in resequencing experiments, we mapped 30 diverse high-coverage Illumina-sequenced samples to GRCh38 with and without contigs containing SV insertions as alternate sequences, and we found these additional sequences recover 6.4% of unmapped reads. For reads mapped within the SV insertion, 25.7% have a better mapping quality, and 18.7% improved by 1,000-fold or more. We reveal 72,964 occurrences of 15,814 unique variants that were not discoverable with the reference sequence alone, and we note that 7% of the insertions contain an SV in at least one sample indicating that there are additional alleles in the population that remain to be discovered. These data provide the framework to construct a canonical human reference and a resource for developing advanced representations capable of capturing allelic diversity. We present a summary of our findings and discuss ideas for revealing variation that was once difficult to ascertain.
Human genomic variations range in size from single nucleotide substitutions to large chromosomal rearrangements. Sequencing technologies tend to be optimized for detecting particular variant types and sizes. Short reads excel at detecting SNVs and small indels, while long or linked reads are typically used to detect larger structural variants or phase distant loci. Long reads are more easily mapped to repetitive regions, but tend to have lower per-base accuracy, making it difficult to call short variants. The PacBio Sequel System produces two main data types: long continuous reads (up to 100 kbp), generated by single passes over a long template, and Circular Consensus Sequence (CCS) reads, generated by calculating the consensus of many sequencing passes over a single shorter template (500 bp to 20 kbp). The long-range information in continuous reads is useful for genome assembly and structural variant detection. The higher base accuracy of CCS effectively detects and phases short variants in single molecules. Recent improvements in library preparation protocols and sequencing chemistry have increased the length, accuracy, and throughput of CCS reads. For the human sample HG002, we collected 28-fold coverage 15 kbp high-fidelity CCS reads with an average read quality above Q20 (99% accuracy). The length and accuracy of these reads allow us to detect SNVs, indels, and structural variants not only in the Genome in a Bottle (GIAB) high confidence regions, but also in segmental duplications, HLA loci, and clinically relevant “difficult-to-map” genes. As with continuous long reads, we call structural variants at 90.0% recall compared to the GIAB structural variant benchmark “truth” set, with the added advantages of base pair resolution for variant calls and improved recall at compound heterozygous loci. With minimap2 alignments, GATK4 HaplotypeCaller variant calls, and simple variant filtration, we have achieved a SNP F-Score of 99.51% and an INDEL F-Score of 80.10% against the GIAB short variant benchmark “truth” set, in addition to calling variants outside of the high confidence region established by GIAB using previous technologies. With the long-range information available in 15 kbp reads, we applied the read-backed phasing tool WhatsHap to generate phase blocks with a mean length of 65 kbp across the entire genome. Using an alignment-based approach, we typed all major MHC class I and class II genes to at least 3-field precision. This new data type has the potential to expand the GIAB high confidence regions and “truth” benchmark sets to many previously difficult-to-map genes and allow a single sequencing protocol to address both short variants and large structural variants.
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 healthy 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. 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. To comprehensively detect large variants in human genomes, we 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. Together with insertion, deletion, inversion, duplication, and translocation calling from spanning reads, this allows pbsv to comprehensively detect large variants from a single data type.
Andrew Carroll, Director of Science at DNAnexus, presents how to greatly improve the accuracy of SV-calling by using long-read PacBio sequencing and fast and easy-to-run cloud-optimized apps like PBHoney, Parliament,…
Fritz Sedlazeck, a postdoc at Johns Hopkins University, describes his structural variant detection tool Sniffles in this poster from AGBT 2016. Included: examples of structural variants that could not be…
At AGBT 2017, Mike Schatz from Johns Hopkins University and Cold Spring Harbor Laboratory presented data from sequencing, assembling, and analyzing personalized, phased diploid genomes with either Illumina, 10x Genomics,…