AGBT 2013 Presentation Slides: Cold Spring Harbor Laboratory’s Michael Schatz presented strategies for de novo assembly of crop genomes with PacBio technolgy.
Advances in sequence consensus and clustering algorithms for effective de novo assembly and haplotyping applications.
One of the major applications of DNA sequencing technology is to bring together information that is distant in sequence space so that understanding genome structure and function becomes easier on a large scale. The Single Molecule Real Time (SMRT) Sequencing platform provides direct sequencing data that can span several thousand bases to tens of thousands of bases in a high-throughput fashion. In contrast to solving genomic puzzles by patching together smaller piece of information, long sequence reads can decrease potential computation complexity by reducing combinatorial factors significantly. We demonstrate algorithmic approaches to construct accurate consensus when the differences between reads are dominated by insertions and deletions. High-performance implementations of such algorithms allow more efficient de novo assembly with a pre-assembly step that generates highly accurate, consensus-based reads which can be used as input for existing genome assemblers. In contrast to recent hybrid assembly approach, only a single ~10 kb or longer SMRTbell library is necessary for the hierarchical genome assembly process (HGAP). Meanwhile, with a sensitive read-clustering algorithm with the consensus algorithms, one is able to discern haplotypes that differ by less than 1% different from each other over a large region. One of the related applications is to generate accurate haplotype sequences for HLA loci. Long sequence reads that can cover the whole 3 kb to 4 kb diploid genomic regions will simplify the haplotyping process. These algorithms can also be applied to resolve individual populations within mixed pools of DNA molecules that are similar to each, e.g., by sequencing viral quasi-species samples.
Using whole exome sequencing and bacterial pathogen sequencing to investigate the genetic basis of pulmonary non-tuberculous mycobacterial infections.
Pulmonary non-tuberculous mycobacterial (PNTM) infections occur in patients with chronic lung disease, but also in a distinct group of elderly women without lung defects who share a common body morphology: tall and lean with scoliosis, pectus excavatum, and mitral valve prolapse. In order to characterize the human host susceptibility to PNTM, we performed whole exome sequencing (WES) of 44 individuals in extended families of patients with active PNTM as well as 55 additional unrelated individuals with PNTM. This unique collection of familial cohorts in PNTM represents an important opportunity for a high yield search for genes that regulate mucosal immunity. An average of 58 million 100bp paired-end Illumina reads per exome were generated and mapped to the hg19 reference genome. Following variant detection and classification, we identified 58,422 potentially high-impact SNPs, 97.3% of which were missense mutations. Segregating variants using the family pedigrees as well as comparisons to the unrelated individuals identified multiple potential variants associated with PNTM. Validations of these candidate variants in a larger PNTM cohort are underway. In addition to WES, we sequenced the genomes of 52 mycobacterial isolates, including 9 from these PNTM patients, to integrate host PNTM susceptibility with mycobacterial genotypes and gain insights into the key factors involved in this devastating disease. These genomes were sequenced using a combination of 454, Illumina, and PacBio platforms and assembled using multiple genome assemblers. The resulting genome sequences were used to identify mycobacterial genotypes associated with virulence, invasion, and drug resistance.
Single Molecule, Real-Time (SMRT) Sequencing holds promise for addressing new frontiers in large genome complexities, such as long, highly repetitive, low-complexity regions and duplication events, and differentiating between transcript isoforms that are difficult to resolve with short-read technologies. We present solutions available for both reference genome improvement (>100 MB) and transcriptome research to best leverage long reads that have exceeded 20 Kb in length. Benefits for these applications are further realized with consistent use of size-selection of input sample using the BluePippin™ device from Sage Science. Highlights from our genome assembly projects using the latest P5-C3 chemistry on model organisms will be shared. Assembly contig N50 have exceeded 6 Mb and we observed longest contig exceeding 12.5 Mb with an average base quality of QV50. Additionally, the value of long, intact reads to provide a no-assembly approach to investigate transcript isoforms using our Iso-Seq Application will be presented.
A comparison of assemblers and strategies for complex, large-genome sequencing with PacBio long reads.
PacBio sequencing holds promise for addressing large-genome complexities, such as long, highly repetitive, low-complexity regions and duplication events that are difficult to resolve with short-read technologies. Several strategies, with varying outcomes, are available for de novo sequencing and assembling of larger genomes. Using a diploid fungal genome, estimated to be ~80 Mb in size, as the basis dataset for comparison, we highlight assembly options when using only PacBio sequencing or a combined strategy leveraging data sets from multiple sequencing technologies. Data generated from SMRT Sequencing was subjected to assembly using different large-genome assemblers, and comparisons of the results will be shown. These include results generated with HGAP, Celera Assembler, MIRA, PBJelly, and other assembly tools currently in development. Improvements observed include a near 50% reduction in the number of contigs coupled with at least a doubling of contig N50 size in genome assemblies incorporating SMRT Sequencing data. We further show how incorporating long reads also highlights new challenges and missed insights of short-read assemblies arising from heterozygosity inherent in multiploid genomes.
Third generation single molecule sequencing technology from Pacific Biosciences, Moleculo, Oxford Nanopore, and other companies are revolutionizing genomics by enabling the sequencing of long, individual molecules of DNA and RNA. One major advantage of these technologies over current short read sequencing is the ability to sequence much longer molecules, thousands or tens of thousands of nucleotides instead of mere hundreds. This capacity gives researchers substantially greater power to probe into microbial, plant, and animal genomes, but it remains unknown on how to best use these data. To answer this, we systematically evaluated the human genome and 25 other important genomes across the tree of life ranging in size from 1Mbp to 3Gbp in an attempt to answer how long the reads need to be and how much coverage is necessary to completely assemble their chromosomes with single molecule sequencing. We also present a novel error correction and assembly algorithm using a combination of PacBio and pre-assembled Illumina sequencing. This new algorithm greatly outperforms other published hybrid algorithms.
Single Molecule, Real-Time (SMRT) Sequencing holds promise for addressing new frontiers to understand molecular mechanisms in evolution and gain insight into adaptive strategies. With read lengths exceeding 10 kb, we are able to sequence high-quality, closed microbial genomes with associated plasmids, and investigate large genome complexities, such as long, highly repetitive, low-complexity regions and multiple tandem-duplication events. Improved genome quality, observed at 99.9999% (QV60) consensus accuracy, and significant reduction of gap regions in reference genomes (up to and beyond 50%) allow researchers to better understand coding sequences with high confidence, investigate potential regulatory mechanisms in noncoding regions, and make inferences about evolutionary strategies that are otherwise missed by the coverage biases associated with short- read sequencing technologies. Additional benefits afforded by SMRT Sequencing include the simultaneous capability to detect epigenomic modifications and obtain full-length cDNA transcripts that obsolete the need for assembly. With direct sequencing of DNA in real-time, this has resulted in the identification of numerous base modifications and motifs, which genome-wide profiles have linked to specific methyltransferase activities. Our new offering, the Iso-Seq Application, allows for the accurate differentiation between transcript isoforms that are difficult to resolve with short-read technologies. PacBio reads easily span transcripts such that both 5’/3’ primers for cDNA library generation and the poly-A tail are observed. As such, exon configuration and intron retention events can be analyzed without ambiguity. This technological advance is useful for characterizing transcript diversity and improving gene structure annotations in reference genomes. We review solutions available with SMRT Sequencing, from targeted sequencing efforts to obtaining reference genomes (>100 Mb). This includes strategies for identifying microsatellites and conducting phylogenetic comparisons with targeted gene families. We highlight how to best leverage our long reads that have exceeded 20 kb in length for research investigations, as well as currently available bioinformatics strategies for analysis. Benefits for these applications are further realized with consistent use of size selection of input sample using the BluePippin™ device from Sage Science as demonstrated in our genome improvement projects. Using the latest P5-C3 chemistry on model organisms, these efforts have yielded an observed contig N50 of ~6 Mb, with the longest contig exceeding 12.5 Mb and an average base quality of QV50.
Single Molecule, Real-Time (SMRT) Sequencing provides efficient, streamlined solutions to address new frontiers in plant genomes and transcriptomes. Inherent challenges presented by highly repetitive, low-complexity regions and duplication events are directly addressed with multi- kilobase read lengths exceeding 8.5 kb on average, with many exceeding 20 kb. Differentiating between transcript isoforms that are difficult to resolve with short-read technologies is also now possible. We present solutions available for both reference genome and transcriptome research that best leverage long reads in several plant projects including algae, Arabidopsis, rice, and spinach using only the PacBio platform. Benefits for these applications are further realized with consistent use of size-selection of input sample using the BluePippin™ device from Sage Science. We will share highlights from our genome projects using the latest P5- C3 chemistry to generate high-quality reference genomes with the highest contiguity, contig N50 exceeding 1 Mb, and average base quality of QV50. Additionally, the value of long, intact reads to provide a no-assembly approach to investigate transcript isoforms using our Iso-Seq protocol will be presented for full transcriptome characterization and targeted surveys of genes with complex structures. PacBio provides the most comprehensive assembly with annotation when combining offerings for both genome and transcriptome research efforts. For more focused investigation, PacBio also offers researchers opportunities to easily investigate and survey genes with complex structures.
Heterozygous and highly polymorphic diploid (2n) and higher polyploidy (n > 2) genomes have proven to be very difficult to assemble. One key to the successful assembly and phasing of polymorphic genomics is the very long read length (9-40 kb) provided by the PacBio RS II system. We recently released software and methods that facilitate the assembly and phasing of genomes with ploidy levels equal to or greater than 2n. In an effort to collaborate and spur on algorithm development for assembly and phasing of heterozygous polymorphic genomes, we have recently released sequencing datasets that can be used to test and develop highly polymorphic diploid and polyploidy assembly and phasing algorithms. These data sets include multiple species and ecotypes of Arabidopsis that can be combined to create synthetic in-silico F1 hybrids with varying levels of heterozygosity. Because the sequence of each individual line was generated independently, the data set provides a ‘ground truth’ answer for the expected results allowing the evaluation of assembly algorithms. The sequencing data, assembly of inbred and in-silico heterozygous samples (n=>2) and phasing statistics will be presented. The raw and processed data has been made available to aid other groups in the development of phasing and assembly algorithms.
Second-generation sequencing has brought about tremendous insights into the genetic underpinnings of biology. However, there are many functionally important and medically relevant regions of genomes that are currently difficult or impossible to sequence, resulting in incomplete and fragmented views of genomes. Two main causes are (i) limitations to read DNA of extreme sequence content (GC-rich or AT-rich regions, low complexity sequence contexts) and (ii) insufficient read lengths which leave various forms of structural variation unresolved and result in mapping ambiguities.
Generating de novo reference genome assemblies for non-model organisms is a laborious task that often requires a large amount of data from several sequencing platforms and cytogenetic surveys. By using PacBio sequence data and new library creation techniques, we present a de novo, high quality reference assembly for the goat (Capra hircus) that demonstrates a primarily sequencing-based approach to efficiently create new reference assemblies for Eukaryotic species. This goat reference genome was created using 38 million PacBio P5-C3 reads generated from a San Clemente goat using the Celera Assembler PBcR pipeline with PacBio read self-correction. In order to generate the assembly, corrected and filtered reads were pre-assembled into a consensus model using PBDAGCON, and subsequently assembled using the Celera Assembly version 8.2. We generated 5,902 contigs using this method with a contig N50 size of 2.56 megabases. In order to generate chromosome-sized scaffolds, we used the LACHESIS scaffolding method to identify cis-chromosome Hi-C interactions in order to link contigs together. We then compared our new assembly to the existing goat reference assembly to identify large-scale discrepancies. In our comparison, we identified 247 disagreements between the two assemblies consisting of 123 inversions and 124 chromosome-contig relocations. The high quality of this data illustrates how this methodology can be used to efficiently generate new reference genome assemblies without the use of expensive fluorescent cytometry or large quantities of data from multiple sequencing platforms.
The goat (Capra hircus) remains an important livestock species due to the species’ ability to forage and provide milk, meat and wool in arid environments. The current goat reference assembly and annotation borrows heavily from other loosely related livestock species, such as cattle, and may not reflect the unique structural and functional characteristics of the species. We present preliminary data from a new de novo reference assembly for goat that primarily utilizes 38 million PacBio P5-C3 reads generated from an inbred San Clemente goat. This assembly consists of only 5,902 contigs with a contig N50 size of 2.56 megabases which were grouped into scaffolds using cis-chromosome associations generated by the analysis of Hi-C sequence reads. To provide accurate functional genetic annotation, we utilized existing RNA-seq data and generated new data consisting of over 784 million reads from a combination of 27 different developmental timepoints/tissues. This dataset provides a tangible improvement over existing goat genomics resources by correcting over 247 misassemblies in the current goat reference genome and by annotating predicted gene models with actual expressed transcript data. Our goal is to provide a high quality resource to researchers to enable future genomic selection and functional prediction within the field of goat genomics.
De novo assembly of a complex panicoid grass genome using ultra-long PacBio reads with P6C4 chemistry
Drought is responsible for much of the global losses in crop yields and understanding how plants naturally cope with drought stress is essential for breeding and engineering crops for the changing climate. Resurrection plants desiccate to complete dryness during times of drought, then “come back to life” once water is available making them an excellent model for studying drought tolerance. Understanding the molecular networks governing how resurrection plants handle desiccation will provide targets for crop engineering. Oropetium thomaeum (Oro) is a resurrection plant that also has the smallest known grass genome at 250 Mb compared to Brachypodium distachyon (300 Mb) and rice (350 Mb). Plant genomes, especially grasses, have complex repeat structures such as telomeres, centromeres, and ribosomal gene cassettes, and high heterozygosity, which makes them difficult to assembly using short read next generation sequencing technologies. Ultra-long PacBio reads using the new P6C4 chemistry and the latest 15kb Blue Pippin size-selection protocol to generate 20kb insert libraries that yielded an average read length of 12kb providing ~72X coverage, and 10X coverage with reads over 20kb. The HGAP assembly covers 98% of the genome with a contig N50 of 2.4 Mb, which makes it one of the highest quality and most complete plant genomes assembled to date. Oro has a compact genome structure compared to other grasses with only 16% repeat sequences but has very good collinearity with other grasses. Understanding the genomic mechanisms of extreme desiccation tolerance in resurrection plants like Oro will provide insights for engineering and intelligent breeding of improved food, fuel, and fiber crops.
Goat is an important source of milk, meat, and fiber, especially in developing countries. An advantage of goats as livestock is the low maintenance requirements and high adaptability compared to other milk producers. The global population of domestic goats exceeds 800 million. In Africa, goat production is characterized by low productivity levels, and attempts to introduce more productive breeds have met with poor success due in part to nutritional constraints. It has been suggested that incorporation of selective breeding within the herds adapted for survival could represent one approach to improving food security across Africa. A recently produced genome assembly of a Chinese Yunnan breed goat, based on 192 Gb of short reads across a range of insert sizes from 180 bp to 20 kb, reported a contig N50 of 18.7 kb. The scaffold N50 was improved from 2.2 Mb to 3.1 Mb by addition of fosmid end sequence, with an estimated 140 million Ns in gaps and 91% coverage. The assembly has proven somewhat problematic for pursuing genome-wide association analysis with SNP arrays, apparently due in part to errors in ordering of markers using the draft genome. In order to provide a higher quality assembly, we sequenced a highly inbred, San Clemente breed goat genome using 458 SMRT cells on the Pacific Biosciences platform. These cells generated 193.5 Gbases of sequence after processing into subreads, with mean 5110 bases and max subread length of 40.5 kb. This sequence data generated an assembly using the recently reported MHAP error correction approach and Celera Assembler v8.2. The contig N50 was 2.5 Mb, with the largest contig spanning 19.5 Mb. Additional characteristics of the assembly will be presented.
Significant advances in bioinformatics tool development have been made to more efficiently leverage and deliver high-quality genome assemblies with PacBio long-read data. Current data throughput of SMRT Sequencing delivers average read lengths ranging from 10-15 kb with the longest reads exceeding 40 kb. This has resulted in consistent demonstration of a minimum 10-fold improvement in genome assemblies with contig N50 in the megabase range compared to assemblies generated using only short- read technologies. This poster highlights recent advances and resources available for advanced bioinformaticians and developers interested in the current state-of-the-art large genome solutions available as open-source code from PacBio and third-party solutions, including HGAP, MHAP, and ECTools. Resources and tools available on GitHub are reviewed, as well as datasets representing major model research organisms made publically available for community evaluation or interested developers.