2015 SMRT Informatics Developers Conference Presentation Slides: Jason Chin of PacBio highlighted some of the challenges for shotgun assembly while suggesting some potential solutions to obtain diploid assemblies, including the FALCON method.
PacBio bioinformatician, Elizabeth Tseng, reviews the bioinformatics strategies utilizing PacBio long-read sequencing data for isoform sequencing for full-length transcript sequencing without assembly.
2015 SMRT Informatics Developers Conference Presentation Slides: Gene Myers, Ph.D., Founding Director, Systems Biology Center, Max Planck Institute delivered the keynote presentation. He talked about building efficient assemblers, the importance of random error distribution in sequencing data, and resolving tricky repeats with very long reads. He also encouraged developers to release assembly modules openly, and noted that data should be straightforward to parse since sharing data interfaces is easier than sharing software interfaces.
MHC class I and II genes are critically monitored by high-resolution sequencing for organ transplant decisions due to their role in GVHD. Their direct or linkage-based causal association, have increased their prominence as targets for drug sensitivity, autoimmune, cancer and infectious disease research. Monitoring HLA genes can however be tricky due to their highly polymorphic nature. Allele-level resolution is thus strongly preferred. However, most studies were historically focused on peptide binding domains of the HLA genes, due to technological challenges. As a result knowledge about the functional role of polymorphisms outside of exons 2 and 3 of HLA genes was rather limited. There are also relatively few full-length gene references currently available in the IMGT HLA database. This made it difficult to quickly adopt high-throughput reference-reliant methods for allele-level HLA sequencing. Increasing awareness regarding role of regulatory region polymorphisms of HLA genes in disease association1, nonetheless have brought about a revolution in full-length HLA gene sequencing. Researchers are now exploring ways to obtain complete information for HLA genes and integrate it with the current HLA database so it can be interpreted used by clinical researchers. We have explored advantages of SMRT Sequencing to obtain fully phased, allele-specific sequences of HLA class I and II genes for 96 samples using completely De novo consensus generation approach for imputation-free 4-field typing. With long read lengths (average >10 kb) and consensus accuracy exceeding 99.999% (Q50), a comprehensive snapshot of variants in exons, introns and UTRs could be obtained for spectrum of polymorphisms in phase across SNP-poor regions. Such information can provide invaluable insights in future causality association and population diversity research.
Reference quality de novo genome assemblies were once solely the domain of large, well-funded genome projects. While next-generation short read technology removed some of the cost barriers, accurate chromosome-scale assembly remains a real challenge. Here we present efforts to de novo assemble the goat (Capra hircus) genome. Through the combination of single-molecule technologies from Pacific Biosciences (sequencing) and BioNano Genomics (optical mapping) coupled with high-throughput chromosome conformation capture sequencing (Hi-C), an inbred San Clemente goat genome has been sequenced and assembled to a high degree of completeness at a relatively modest cost. Starting with 38 million PacBio reads, we integrated the MinHash Alignment Process (MHAP) with the Celera Assembler (CA) to produce an assembly composed of 3110 contigs with a contig N50 size of 4.7 Mb. This assembly was scaffolded with BioNano genome maps derived from a single IrysChip into 333 scaffolds with an N50 of 23.1 Mb including the complete scaffolding of chromosome 20. Finally, cis-chromosome associations were determined by Hi-C, yielding complete reconstruction of all autosomes into single scaffolds with a final N50 of 91.7 Mb. We hope to demonstrate that our methods are not only cost effective, but improve our ability to annotate challenging genomic regions such as highly repetitive immune gene clusters.
Outside of the simplest cases (haploid, bacteria, or inbreds), genomic information is not carried in a single reference per individual, but rather has higher ploidy (n=>2) for almost all organisms. The existence of two or more highly related sequences within an individual makes it extremely difficult to build high quality, highly contiguous genome assemblies from short DNA fragments. Based on the earlier work on a polyploidy aware assembler, FALCON ( https://github.com/PacificBiosciences/FALCON) , we developed new algorithms and software (“FALCON-unzip”) for de novo haplotype reconstructions from SMRT Sequencing data. We generate two datasets for developing the algorithms and the prototype software: (1) whole genome sequencing data from a highly repetitive diploid fungal (Clavicorona pyxidata) and (2) whole genome sequencing data from an F1 hybrid from two inbred Arabidopsis strains: Cvi-0 and Col-0. For the fungal genome, we achieved an N50 of 1.53 Mb (of the 1n assembly contigs) of the ~42 Mb 1n genome and an N50 of the haplotigs (haplotype specific contigs) of 872 kb from a 95X read length N50 ~16 kb dataset. We found that ~ 45% of the genome was highly heterozygous and ~55% of the genome was highly homozygous. We developed methods to assess the base-level accuracy and local haplotype phasing accuracy of the assembly with short-read data from the Illumina® platform. For the ArabidopsisF1 hybrid genome, we found that 80% of the genome could be separated into haplotigs. The long range accuracy of phasing haplotigs was evaluated by comparing them to the assemblies from the two inbred parental lines. We show that a more complete view of all haplotypes could provide useful biological insights through improved annotation, characterization of heterozygous variants of all sizes, and resolution of differential allele expression. The current Falcon-Unzip method will lead to understand how to solve more difficult polyploid genome assembly problems and improve the computational efficiency for large genome assemblies. Based on this work, we can develop a pipeline enabling routinely assemble diploid or polyploid genomes as haplotigs, representing a comprehensive view of the genomes that can be studied with the information at hand.
Outside of the simplest cases (haploid, bacteria, or inbreds), genomic information is not carried in a single reference per individual, but rather has higher ploidy (n=>2) for almost all organisms. The existence of two or more highly related sequences within an individual makes it extremely difficult to build high quality, highly contiguous genome assemblies from short DNA fragments. Based on the earlier work on a polyploidy aware assembler, FALCON (https://github.com/PacificBiosciences/FALCON), we developed new algorithms and software (FALCON-unzip) for de novo haplotype reconstructions from SMRT Sequencing data. We apply the algorithms and the prototype software for (1) a highly repetitive diploid fungal genome (Clavicorona pyxidata) and (2) an F1 hybrid from two inbred Arabidopsis strains: CVI-0 and COL-0. For the fungal genome, we achieved an N50 of 1.53 Mb (of the 1n assembly contigs) of the ~42 Mb 1n genome and an N50 of the haplotigs of 872 kb from a 95X read length N50 ~16 kb dataset. We found that ~ 45% of the genome was highly heterozygous and ~55% of the genome was highly homozygous. We developed methods to assess the base-level accuracy and local haplotype phasing accuracy of the assembly with short-read data from the Illumina platform. For the Arabidopsis F1 hybrid genome, we found that 80% of the genome could be separated into haplotigs. The long range accuracy of phasing haplotigs was evaluated by comparing them to the assemblies from the two inbred parental lines. We show that a more complete view of all haplotypes could provide useful biological insights through improved annotation, characterization of heterozygous variants of all sizes, and resolution of differential allele expression. Finally, we applied this method to WGS human data sets to demonstrate the potential for resolving complicated, medically-relevant genomic regions.
Scientists who require confident resolution of heterogeneous populations across complex regions have been unable to transition to short-read sequencing methods. They continue to depend on Sanger Sequencing despite its cost and time inefficiencies. Here we present a new redesigned algorithm that allows the generation of circular consensus sequences (CCS) from individual SMRT Sequencing reads. With this new algorithm, dubbed CCS2, it is possible to reach arbitrarily high quality across longer insert lengths at a lower cost and higher throughput than Sanger Sequencing. We apply this new algorithm, dubbed CCS2, to the characterization of the HIV-1 K103N drug-resistance associated mutation, which is both important clinically, and represents a challenge due to regional sequence context. A mutation was introduced into the 3rd position of amino acid position 103 (A>C substitution) of the RT gene on a pNL4-3 backbone by site-directed mutagenesis. Regions spanning ~1,300 bp were PCR amplified from both the non-mutated and mutant (K103N) plasmids, and were sequenced individually and as a 50:50 mixture. Sequencing data were analyzed using the new CCS2 algorithm, which uses a fully-generative probabilistic model of our SMRT Sequencing process to polish consensus sequences to arbitrarily high accuracy. This result, previously demonstrated for multi-molecule consensus sequences with the Quiver algorithm, is made possible by incorporating per-Zero Mode Waveguide (ZMW) characteristics, thus accounting for the intrinsic changes in the sequencing process that are unique to each ZMW. With CCS2, we are able to achieve a per-read empirical quality of QV30 with 19X coverage. This yields ~5000 1.3 kb consensus sequences with a collective empirical quality of ~QV40. Additionally, we demonstrate a 0% miscall rate in both unmixed samples, and estimate a 48:52% frequency for the K103N mutation in the mixed sample, consistent with data produced by orthogonal platforms.
For highly complex and large genomes, a well-annotated genome may be computationally challenging and costly, yet the study of alternative splicing events and gene annotations usually rely on the existence of a genome. Long-read sequencing technology provides new opportunities to sequence full-length cDNAs, avoiding computational challenges that short read transcript assembly brings. The use of single molecule, real-time sequencing from PacBio to sequence transcriptomes (the Iso-Seq method), which produces de novo, high-quality, full-length transcripts, has revealed an astonishing amount of alternative splicing in eukaryotic species. With the Iso-Seq method, it is now possible to reconstruct the transcribed regions of the genome using just the transcripts themselves. We present Cogent, a tool for finding gene families and reconstructing the coding genome in the absence of a high-quality reference genome. Cogent uses k-mer similarities to first partition the transcripts into different gene families. Then, for each gene family, the transcripts are used to build a splice graph. Cogent identifies bubbles resulting from sequencing errors, minor variants, and exon skipping events, and attempts to resolve each splice graph down to the minimal set of reconstructed contigs. We apply Cogent to the Iso-Seq data for spinach, Spinacia oleracea, for which there is also a PacBio-based draft genome to validate the reconstruction. The Iso-Seq dataset consists of 68,263 fulllength, Quiver-polished transcript sequences ranging from 528 bp to 6 kbp long (mean: 2.1 kbp). Using the genome mapping as ground truth, we found that 95% (8045/8446) of the Cogent gene families found corresponded to a single genomic loci. For families that contained multiple loci, they were often homologous genes that would be categorized as belonging to the same gene family. Coding genome reconstruction was then performed individually for each gene family. A total of 86% (7283/8446) of the gene families were resolved to a single contig by Cogent, and was validated to be also a single contig in the genome. In 59 cases, Cogent reconstructed a single contig, however the contig corresponded to 2 or more loci in the genome, suggesting possible scaffolding opportunities. In 24 cases, the transcripts had no hits to the genome, though Pfam and BLAST searches of the transcripts show that they were indeed coding, suggesting that the genome is missing certain coding portions. Given the high quality of the spinach genome, we were not surprised to find that Cogent only minorly improved the genome space. However the ability of Cogent to accurately identify gene families and reconstruct the coding genome in a de novo fashion shows that it will be extremely powerful when applied to datasets for which there is no or low-quality reference genome.
The killer immunoglobulin-like receptors (KIR) genes belong to the immunoglobulin superfamily and are widely studied due to the critical role they play in coordinating the innate immune response to infection and disease. Highly accurate, contiguous, long reads, like those generated by SMRT Sequencing, when combined with target-enrichment protocols, provide a straightforward strategy for generating complete de novo assembled KIR haplotypes. We have explored two different methods to capture the KIR region; one applying the use of fosmid clones and one using Nimblegen capture.
The long reads, random error, and unbiased sampling of SMRT Sequencing enables high quality, de novo assembly of the human genome. PacBio long reads are capable of resolving genomic variations at all size scales, including SNPs, insertions, deletions, inversions, translocations, and repeat expansions, all of which are both important in understanding the genetic basis for human disease, and difficult to access via other technologies. In demonstration of this, we report a new high-quality, diploid-aware de novo assembly of Craig Venter’s well-studied genome.
Scientists who require confident resolution of heterogeneous populations across complex regions have been unable to transition to short-read sequencing methods. They continue to depend on Sanger sequencing despite its cost and time inefficiencies. Here we present a new redesigned algorithm that allows the generation of circular consensus sequences (CCS) from individual SMRT Sequencing reads. With this new algorithm, dubbed CCS2, it is possible to reach high quality across longer insert lengths at a lower cost and higher throughput than Sanger sequencing. We applied CCS2 to the characterization of the HIV-1 K103N drug-resistance associated mutation in both clonal and patient samples. This particular DRAM has previously proved to be clinically relevant, but challenging to characterize due to regional sequence context. First, a mutation was introduced into the 3rd position of amino acid position 103 (A>C substitution) of the RT gene on a pNL4-3 backbone by site-directed mutagenesis. Regions spanning ~1.3 kb were PCR amplified from both the non-mutated and mutant (K103N) plasmids, and were sequenced individually and as a 50:50 mixture. Additionally, the proviral reservoir of a subject with known dates of virologic failure of an Efavirenz-based regimen and with documented emergence of drug resistant (K103N) viremia was sequenced at several time points as a proof-of-concept study to determine the kinetics of retention and decay of K103N.Sequencing data were analyzed using the new CCS2 algorithm, which uses a fully-generative probabilistic model of our SMRT Sequencing process to polish consensus sequences to high accuracy. With CCS2, we are able to achieve a per-read empirical quality of QV30 (99.9% accuracy) at 19X coverage. A total of ~5000 1.3 kb consensus sequences with a collective empirical quality of ~QV40 (99.99%) were obtained for each sample. We demonstrate a 0% miscall rate in both unmixed control samples, and estimate a 48:52 frequency for the K103N mutation in the mixed (50:50) plasmid sample, consistent with data produced by orthogonal platforms. Additionally, the K103N escape variant was only detected in proviral samples from time points subsequent (19%) to the emergence of drug resistant viremia. This tool might be used to monitor the HIV reservoir for stable evolutionary changes throughout infection.
Alzheimer’s disease (AD) is a devastating neurodegenerative disease that is genetically complex. Although great progress has been made in identifying fully penetrant mutations in genes such as APP, PSEN1 and PSEN2 that cause early-onset AD, these still represent a very small percentage of AD cases. Large-scale, genome-wide association studies (GWAS) have identified at least 20 additional genetic risk loci for the more common form of late-onset AD. However, the identified SNPs are typically not the actual risk variants, but are in linkage disequilibrium with the presumed causative variant (Van Cauwenberghe C, et al., The genetic landscape of Alzheimer disease: clinical implications and perspectives. Genet Med 2015;18:421-430). Long-read sequencing together with hybrid-capture targeting technologies provides a powerful combination to target candidate genes/transcripts of interest. Shearing the genomic DNA to ~5 kb fragments and then capturing with probes that span the whole gene(s) of interest can provide uniform coverage across the entire region, identifying variants and allowing for phasing into two haplotypes. Furthermore, capturing full-length cDNA from the same sample using the same capture probes can also provide an understanding of isoforms that are generated and allow them to be assigned to their corresponding haplotype. Here we present a method for capturing genomic DNA and cDNA from an AD sample using a panel of probes targeting approximately 20 late-onset AD candidate genes which includes CLU, ABCA7, CD33, TREM2, TOMM40, PSEN2, APH1 and BIN1. By combining xGen® Lockdown® probes with SMRT Sequencing, we provide completely sequenced candidate genes as well as their corresponding transcripts. In addition, we are also able to evaluate structural variants that due to their size, repetitive nature, or low sequence complexity have been un-sequenceable using short-read technologies.
Genes associated with several neurological disorders have been shown to be highly polymorphic. Targeted sequencing of these genes using NGS technologies is a powerful way to increase the cost-effectiveness of variant discovery and detection. However, for a comprehensive view of these target genes, it is necessary to have complete and uniform coverage across regions of interest. Unfortunately, short-read sequencing technologies are not ideal for these types of studies as they are prone to mis-mapping and often fail to span repetitive regions. Targeted sequencing with PacBio long reads provides the unique advantage of single-molecule observations of complex genomic regions. PacBio long reads not only provide continuous sequence data though polymorphic or repetitive regions, but also have no GC bias. Here we describe the characterization of the poly-T locus in TOMM40, a gene known to be associated with progression to Alzheimer’s, using PacBio long reads. Probes were designed to capture a 20 kb region comprising the TOMM40 and ApoE genes. Target regions were captured in multiple cell lines and sequencing libraries made using standard sample preparation methods. We will present our results on the poly-T structural variants that we observed in TOMM40 in these cell lines. We will also present our results on probe design optimization and barcoding strategies for a cost-effective solution.
Many applications of high throughput sequencing rely on the availability of an accurate reference genome. Errors in the reference genome assembly increase the number of false-positives in downstream analyses. Recently, we have shown that over 33% of the current pig reference genome, Sscrofa10.2, is either misassembled or otherwise unreliable for genomic analyses. Additionally, ~10% of the bases in the assembly are Ns in gaps of an arbitrary size. Thousands of highly fragmented contigs remain unplaced and many genes are known to be missing from the assembly. Here we present a new assembly of the pig genome, Sscrofa11, assembled using 65X PacBio sequencing from T.J. Tabasco, the same Duroc sow used in the assembly of Sscrofa10.2. The PacBio reads were assembled using the Falcon assembly pipeline resulting in 3,206 contigs with an initial contig N50 of 14.5Mb. We used Sscrofa10.2 as a template to scaffold the PacBio contigs, under the assumption that its gross structure is correct, and used PBJelly to fill gaps. Additional gaps were filled using large, sequenced BACs from the original assembly. Following gap filling, the assembly has substantially improved contiguity and contains more sequence than the Sscrofa10.2 assembly. Arrow and Pilon were used to polish the assembly. The contig N50 is now 58.5Mb with 103 gaps remaining. By comparing regions of the two assemblies we show that regions with structural abnormalities we identified in Sscrofa10.2 are resolved in the new PacBio assembly.
If you have a question, need to check the status of an order, or are interested in purchasing an instrument, we're here to help.