Short-read sequencing has enabled the de novo assembly of several individual human genomes, but with inherent limitations in characterizing repeat elements. Here we sequence a Chinese individual HX1 by single-molecule real-time (SMRT) long-read sequencing, construct a physical map by NanoChannel arrays and generate a de novo assembly of 2.93?Gb (contig N50: 8.3?Mb, scaffold N50: 22.0?Mb, including 39.3?Mb N-bases), together with 206?Mb of alternative haplotypes. The assembly fully or partially fills 274 (28.4%) N-gaps in the reference genome GRCh38. Comparison to GRCh38 reveals 12.8?Mb of HX1-specific sequences, including 4.1?Mb that are not present in previously reported Asian genomes. Furthermore, long-read sequencing of the transcriptome reveals novel spliced genes that are not annotated in GENCODE and are missed by short-read RNA-Seq. Our results imply that improved characterization of genome functional variation may require the use of a range of genomic technologies on diverse human populations.
In recent years long-read technologies have moved from being a niche and specialist field to a point of relative maturity likely to feature frequently in the genomic landscape. Analogous to next generation sequencing, the cost of sequencing using long-read technologies has materially dropped whilst the instrument throughput continues to increase. Together these changes present the prospect of sequencing large numbers of individuals with the aim of fully characterizing genomes at high resolution. In this article, we will endeavour to present an introduction to long-read technologies showing: what long reads are; how they are distinct from short reads; why long reads are useful and how they are being used. We will highlight the recent developments in this field, and the applications and potential of these technologies in medical research, and clinical diagnostics and therapeutics.
Performing sequence alignment to identify structural variants, such as large deletions, from genome sequencing data is a fundamental task, but current methods are far from perfect. The current practice is to independently align each DNA read to a reference genome. We show that the propensity of genomic rearrangements to accumulate in repeat-rich regions imposes severe ambiguities in these alignments, and consequently on the variant calls-with current read lengths, this affects more than one third of known large deletions in the C. Venter genome. We present a method to jointly align reads to a genome, whereby alignment ambiguity of one read can be disambiguated by other reads. We show this leads to a significant improvement in the accuracy of identifying large deletions (=20 bases), while imposing minimal computational overhead and maintaining an overall running time that is at par with current tools. A software implementation is available as an open-source Python program called JRA at https://bitbucket.org/jointreadalignment/jra-src.
Genomes mutate and evolve in ways simple (substitution or deletion of bases) and complex (e.g. chromosome shattering). We do not fully understand what types of complex mutation occur, and we cannot routinely characterize arbitrarily-complex mutations in a high-throughput, genome-wide manner. Long-read DNA sequencing methods (e.g. PacBio, nanopore) are promising for this task, because one read may encompass a whole complex mutation. We describe an analysis pipeline to characterize arbitrarily-complex ‘local’ mutations, i.e. intrachromosomal mutations encompassed by one DNA read. We apply it to nanopore and PacBio reads from one human cell line (NA12878), and survey sequence rearrangements, both real and artifactual. Almost all the real rearrangements belong to recurring patterns or motifs: the most common is tandem multiplication (e.g. heptuplication), but there are also complex patterns such as localized shattering, which resembles DNA damage by radiation. Gene conversions are identified, including one between hemoglobin gamma genes. This study demonstrates a way to find intricate rearrangements with any number of duplications, deletions, and repositionings. It demonstrates a probability-based method to resolve ambiguous rearrangements involving highly similar sequences, as occurs in gene conversion. We present a catalog of local rearrangements in one human cell line, and show which rearrangement patterns occur.
Reproducible integration of multiple sequencing datasets to form high-confidence SNP, indel, and reference calls for five human genome reference materials
Benchmark small variant calls from the Genome in a Bottle Consortium (GIAB) for the CEPH/HapMap genome NA12878 (HG001) have been used extensively for developing, optimizing, and demonstrating performance of sequencing and bioinformatics methods. Here, we develop a reproducible, cloud-based pipeline to integrate multiple sequencing datasets and form benchmark calls, enabling application to arbitrary human genomes. We use these reproducible methods to form high-confidence calls with respect to GRCh37 and GRCh38 for HG001 and 4 additional broadly-consented genomes from the Personal Genome Project that are available as NIST Reference Materials. These new genomes’ broad, open consent with few restrictions on availability of samples and data is enabling a uniquely diverse array of applications. Our new methods produce 17% more high-confidence SNPs, 176% more indels, and 12% larger regions than our previously published calls. To demonstrate that these calls can be used for accurate benchmarking, we compare other high-quality callsets to ours (e.g., Illumina Platinum Genomes), and we demonstrate that the majority of discordant calls are errors in the other callsets, We also highlight challenges in interpreting performance metrics when benchmarking against imperfect high-confidence calls. We show that benchmarking tools from the Global Alliance for Genomics and Health can be used with our calls to stratify performance metrics by variant type and genome context and elucidate strengths and weaknesses of a method.
Humans are a diploid species that inherit one set of chromosomes paternally and one homologous set of chromosomes maternally. Unfortunately, most human sequencing initiatives ignore this fact in that they do not directly delineate the nucleotide content of the maternal and paternal copies of the 23 chromosomes individuals possess (i.e., they do not ‘phase’ the genome) often because of the costs and complexities of doing so. We compared 11 different widely-used approaches to phasing human genomes using the publicly available ‘Genome-In-A-Bottle’ (GIAB) phased version of the NA12878 genome as a gold standard. The phasing strategies we compared included laboratory-based assays that prepare DNA in unique ways to facilitate phasing as well as purely computational approaches that seek to reconstruct phase information from general sequencing reads and constructs or population-level haplotype frequency information obtained through a reference panel of haplotypes. To assess the performance of the 11 approaches, we used metrics that included, among others, switch error rates, haplotype block lengths, the proportion of fully phase-resolved genes, phasing accuracy and yield between pairs of SNVs. Our comparisons suggest that a hybrid or combined approach that leverages: 1. population-based phasing using the SHAPEIT software suite, 2. either genome-wide sequencing read data or parental genotypes, and 3. a large reference panel of variant and haplotype frequencies, provides a fast and efficient way to produce highly accurate phase-resolved individual human genomes. We found that for population-based approaches, phasing performance is enhanced with the addition of genome-wide read data; e.g., whole genome shotgun and/or RNA sequencing reads. Further, we found that the inclusion of parental genotype data within a population-based phasing strategy can provide as much as a ten-fold reduction in phasing errors. We also considered a majority voting scheme for the construction of a consensus haplotype combining multiple predictions for enhanced performance and site coverage. Finally, we also identified DNA sequence signatures associated with the genomic regions harboring phasing switch errors, which included regions of low polymorphism or SNV density.
IMSindel: An accurate intermediate-size indel detection tool incorporating de novo assembly and gapped global-local alignment with split read analysis.
Insertions and deletions (indels) have been implicated in dozens of human diseases through the radical alteration of gene function by short frameshift indels as well as long indels. However, the accurate detection of these indels from next-generation sequencing data is still challenging. This is particularly true for intermediate-size indels (=50?bp), due to the short DNA sequencing reads. Here, we developed a new method that predicts intermediate-size indels using BWA soft-clipped fragments (unmatched fragments in partially mapped reads) and unmapped reads. We report the performance comparison of our method, GATK, PINDEL and ScanIndel, using whole exome sequencing data from the same samples. False positive and false negative counts were determined through Sanger sequencing of all predicted indels across these four methods. The harmonic mean of the recall and precision, F-measure, was used to measure the performance of each method. Our method achieved the highest F-measure of 0.84 in one sample, compared to 0.56 for GATK, 0.52 for PINDEL and 0.46 for ScanIndel. Similar results were obtained in additional samples, demonstrating that our method was superior to the other methods for detecting intermediate-size indels. We believe that this methodology will contribute to the discovery of intermediate-size indels associated with human disease.
Structural variants (SVs), including small insertion and deletion variants (indels), are challenging to detect through standard alignment-based variant calling methods. Sequence assembly offers a powerful approach to identifying SVs, but is difficult to apply at scale genome-wide for SV detection due to its computational complexity and the difficulty of extracting SVs from assembly contigs. We describe SvABA, an efficient and accurate method for detecting SVs from short-read sequencing data using genome-wide local assembly with low memory and computing requirements. We evaluated SvABA’s performance on the NA12878 human genome and in simulated and real cancer genomes. SvABA demonstrates superior sensitivity and specificity across a large spectrum of SVs and substantially improves detection performance for variants in the 20-300 bp range, compared with existing methods. SvABA also identifies complex somatic rearrangements with chains of short (<1000 bp) templated-sequence insertions copied from distant genomic regions. We applied SvABA to 344 cancer genomes from 11 cancer types and found that short templated-sequence insertions occur in ~4% of all somatic rearrangements. Finally, we demonstrate that SvABA can identify sites of viral integration and cancer driver alterations containing medium-sized (50-300 bp) SVs.© 2018 Wala et al.; Published by Cold Spring Harbor Laboratory Press.
Structural variants (SVs) in human genomes are implicated in a variety of human diseases. Long-read sequencing delivers much longer read lengths than short-read sequencing and may greatly improve SV detection. However, due to the relatively high cost of long-read sequencing, it is unclear what coverage is needed and how to optimally use the aligners and SV callers.In this study, we developed NextSV, a meta-caller to perform SV calling from low coverage long-read sequencing data. NextSV integrates three aligners and three SV callers and generates two integrated call sets (sensitive/stringent) for different analysis purposes. We evaluated SV calling performance of NextSV under different PacBio coverages on two personal genomes, NA12878 and HX1. Our results showed that, compared with running any single SV caller, NextSV stringent call set had higher precision and balanced accuracy (F1 score) while NextSV sensitive call set had a higher recall. At 10X coverage, the recall of NextSV sensitive call set was 93.5 to 94.1% for deletions and 87.9 to 93.2% for insertions, indicating that ~10X coverage might be an optimal coverage to use in practice, considering the balance between the sequencing costs and the recall rates. We further evaluated the Mendelian errors on an Ashkenazi Jewish trio dataset.Our results provide useful guidelines for SV detection from low coverage whole-genome PacBio data and we expect that NextSV will facilitate the analysis of SVs on long-read sequencing data.
Haplotype assembly is the process of assigning the different alleles of the variants covered by mapped sequencing reads to the two haplotypes of the genome of a human individual. Long reads, which are nowadays cheaper to produce and more widely available than ever before, have been used to reduce the fragmentation of the assembled haplotypes since their ability to span several variants along the genome. These long reads are also characterized by a high error rate, an issue which may be mitigated, however, with larger sets of reads, when this error rate is uniform across genome positions. Unfortunately, current state-of-the-art dynamic programming approaches designed for long reads deal only with limited coverages.Here, we propose a new method for assembling haplotypes which combines and extends the features of previous approaches to deal with long reads and higher coverages. In particular, our algorithm is able to dynamically adapt the estimated number of errors at each variant site, while minimizing the total number of error corrections necessary for finding a feasible solution. This allows our method to significantly reduce the required computational resources, allowing to consider datasets composed of higher coverages. The algorithm has been implemented in a freely available tool, HapCHAT: Haplotype Assembly Coverage Handling by Adapting Thresholds. An experimental analysis on sequencing reads with up to 60 × coverage reveals improvements in accuracy and recall achieved by considering a higher coverage with lower runtimes.Our method leverages the long-range information of sequencing reads that allows to obtain assembled haplotypes fragmented in a lower number of unphased haplotype blocks. At the same time, our method is also able to deal with higher coverages to better correct the errors in the original reads and to obtain more accurate haplotypes as a result.HapCHAT is available at http://hapchat.algolab.eu under the GNU Public License (GPL).
Constructing high-quality haplotype-resolved de novo assemblies of diploid genomes is important for revealing the full extent of structural variation and its role in health and disease. Current assembly approaches often collapse the two sequences into one haploid consensus sequence and, therefore, fail to capture the diploid nature of the organism under study. Thus, building an assembler capable of producing accurate and complete diploid assemblies, while being resource-efficient with respect to sequencing costs, is a key challenge to be addressed by the bioinformatics community.We present a novel graph-based approach to diploid assembly, which combines accurate Illumina data and long-read Pacific Biosciences (PacBio) data. We demonstrate the effectiveness of our method on a pseudo-diploid yeast genome and show that we require as little as 50× coverage Illumina data and 10× PacBio data to generate accurate and complete assemblies. Additionally, we show that our approach has the ability to detect and phase structural variants.https://github.com/whatshap/whatshap.Supplementary data are available at Bioinformatics online.
Detection of genomic inversions remains challenging. Many existing methods primarily target inzversions with a non repetitive breakpoint, leaving inverted repeat (IR) mediated non-allelic homologous recombination (NAHR) inversions largely unexplored.We present npInv, a novel tool specifically for detecting and genotyping NAHR inversion using long read sub-alignment of long read sequencing data. We benchmark npInv with other tools in both simulation and real data. We use npInv to generate a whole-genome inversion map for NA12878 consisting of 30 NAHR inversions (of which 15 are novel), including all previously known NAHR mediated inversions in NA12878 with flanking IR less than 7kb. Our genotyping accuracy on this dataset was 94%. We used PCR to confirm the presence of two of these novel inversions. We show that there is a near linear relationship between the length of flanking IR and the minimum inversion size, without inverted repeats.The application of npInv shows high accuracy in both simulation and real data. The results give deeper insight into understanding inversion.
Although numerous algorithms have been developed to identify large chromosomal rearrangements (i.e., genomic structural variants, SVs), there remains a dearth of approaches to evaluate their results. This is significant, as the accurate identification of SVs is still an outstanding problem whereby no single algorithm has been shown to be able to achieve high sensitivity and specificity across different classes of SVs. The method introduced in this chapter, VaPoR, is specifically designed to evaluate the accuracy of SV predictions using third-generation long sequences. This method uses a recurrence approach and collects direct evidence from raw reads thus avoiding computationally costly whole genome assembly. This chapter would describe in detail as how to apply this tool onto different data types.
Copy number variants (CNVs) are known to affect a large portion of the human genome and have been implicated in many diseases. Although whole-genome sequencing (WGS) can help identify CNVs, most analytical methods suffer from limited sensitivity and specificity, especially in regions of low mappability. To address this, we use PopSV, a CNV caller that relies on multiple samples to control for technical variation. We demonstrate that our calls are stable across different types of repeat-rich regions and validate the accuracy of our predictions using orthogonal approaches. Applying PopSV to 640 human genomes, we find that low-mappability regions are approximately 5 times more likely to harbor germline CNVs, in stark contrast to the nearly uniform distribution observed for somatic CNVs in 95 cancer genomes. In addition to known enrichments in segmental duplication and near centromeres and telomeres, we also report that CNVs are enriched in specific types of satellite and in some of the most recent families of transposable elements. Finally, using this comprehensive approach, we identify 3455 regions with recurrent CNVs that were missing from existing catalogs. In particular, we identify 347 genes with a novel exonic CNV in low-mappability regions, including 29 genes previously associated with disease.
Existing benchmark datasets for use in evaluating variant-calling accuracy are constructed from a consensus of known short-variant callers, and they are thus biased toward easy regions that are accessible by these algorithms. We derived a new benchmark dataset from the de novo PacBio assemblies of two fully homozygous human cell lines, which provides a relatively more accurate and less biased estimate of small-variant-calling error rates in a realistic context.