Hybrid Read Sequencing: Applications and Tools

Next-generation sequencing (Illumina) and long read sequencing (PacBio/Oxford Nanopore) platforms each have their own strengths and weaknesses. Recent advances in single molecule real-time (SMRT) and nanopore sequencing technologies have enabled high-quality assemblies from long and inaccurate reads. However, these approaches require high coverage by long reads and remain expensive. On the other hand, the inexpensive short reads technologies produce accurate but fragmented assemblies. Thus, the combination of these techniques led to a new improved approach known as hybrid sequencing.

The hybrid sequencing methods utilize the high-throughput and high-accuracy short read data to correct errors in the long reads. This approach reduces the required amount of costlier long-read sequence data as well as results in more complete assemblies including the repetitive regions. Moreover, PacBio long reads can provide reliable alignments, scaffolds, and rough detections of genomic variants, while short reads refine the alignments, assemblies, and detections to single-nucleotide resolution. The high coverage of short read sequencing data output can also be utilized in downstream quantitative analysis1.

Applications

De novo sequencing

As alternatives to using PacBio sequencing alone for eukaryotic de novo assemblies, error correction strategies using hybrid sequencing have also been developed.

  • Koren et al. developed the PacBio corrected Reads (PBcR) approach for using short reads to correct the errors in long reads2. PBcR has been applied to reads generated by a PacBio RS instrument from phage, prokaryotic and eukaryotic whole genomes, including the previously unsequenced parrot (Melopsittacus undulates) The long-read correction approach, has achieved >99.9% base-call accuracy, leading to substantially better assemblies than non-hybrid sequencing strategies.
  • Also, Bashir et al. used hybrid sequencing data to assemble the two-chromosome genome of a Haitian cholera outbreak strain at >99.9% accuracy in two nearly finished contigs, completely resolving complex regions with clinically relevant structures3.
  • More recently, Goodwin et al. developed an open-source error correction algorithm Nanocorr, specifically for hybrid error correction of Oxford Nanopore reads. They used this error correction method with complementary MiSeq data to produce a highly contiguous and accurate de novo assembly of the Saccharomyces cerevisiae The contig N50 length was more than ten times greater than an Illumina-only assembly with >99.88% consensus identity when compared to the reference. Additionally, this assembly offered a complete representation of the features of the genome with correctly assembled gene cassettes, rRNAs, transposable elements, and other genomic features that were almost entirely absent in the Illumina-only assembly4.

Transcript structure and Gene isoform identification

Besides genome assembly, hybrid sequencing can also be applied to the error correction of PacBio long reads of transcripts. Moreover, it could improve gene isoform identification and abundance estimation.

  • Along with genome assembly, Koren et al. used the PBcR method to identify and confirm full-length transcripts and gene isoforms. As the length of the single-molecule PacBio reads from RNA-Seq experiments is within the size distribution of most transcripts, many PacBio reads represent near full-length transcripts. These long reads can therefore greatly reduce the need for transcript assembly, which requires complex algorithms for short reads and confidently detect alternatively spliced isoforms. However, the predominance of indel errors makes analysis of the raw reads challenging. Both sets of PacBio reads (before and after error-correction) were aligned to the reference genome to determine the ones that matched the exon structure over the entire length of the annotated transcripts. Before correction, only 41 (0.1%) of the PacBio reads exactly matched the annotated exon structure that rose to 12, 065 (24.1%) after correction.
  • Au et al. developed a computational tool called LSC for the correction of raw PacBio reads by short reads5. Applying this tool to 100,000 human brain cerebellum PacBio subreads and 64 million 75-bp Illumina short reads, they reduced the error rate of the long reads by more than 3-fold. In order to identify and quantify full-length gene isoforms, they also developed an Isoform Detection and Prediction tool (IDP), which makes use of TGS long reads and SGS short reads6. Applying LSC and IDP to PacBio long reads and Illumina short reads of the human embryonic stem cell transcriptome, they detected several thousand RefSeq-annotated gene isoforms at full-length. IDP-fusion has also been released for the identification of fusion genes, fusion sites, and fusion gene isoforms from cancer transcriptomes7.
  • Ning et al. developed an analysis method HySeMaFi to decipher gene splicing and estimate the gene isoforms abundance8. Firstly, the method establishes the mapping relationship between the error-corrected long reads and the longest assembled contig in every corresponding gene. According to the mapping data, the true splicing pattern of the genes is detected, followed by quantification of the isoforms.

Personal transcriptomes

Personal transcriptomes are expected to have applications in understanding individual biology and disease, but short read sequencing has been shown to be insufficiently accurate for the identification and quantification of an individual’s genetic variants and gene isoforms9.

  • Using a hybrid sequencing strategy combining PacBio long reads and Illumina short reads, Tilgner et al. sequenced the lymphoblastoid transcriptomes of three family members in order to produce and quantify an enhanced personalized genome annotation. Around 711,000 CCS reads were used to identify novel isoforms, and ∼100 million Illumina paired-end reads were used to quantify the personalized annotation, which cannot be accomplished by the relatively small number of long reads alone. This method produced reads representing all splice sites of a transcript for most sufficiently expressed genes shorter than 3 kb. It provides a de novo approach for determining single-nucleotide variations, which could be used to improve RNA haplotype inference10.

Epigenetics research

  • Beckmann et al. demonstrated the ability of PacBio sequencing to recover previously-discovered epigenetic motifs with m6A and m4C modifications in both low-coverage and high-contamination scenarios11. They were also able to recover many motifs from three mixed strains ( E. coliG. metallireducens, and C. salexigens), even when the motif sequences of the genomes of interest overlap substantially, suggesting that PacBio sequencing is applicable to metagenomics. Their studies infer that hybrid sequencing would be more cost-effective than using PacBio sequencing alone to detect and accurately define k-mers for low proportion genomes.

Hybrid assembly tools

Several algorithms have been developed that can help in the single molecule de novo assembly of genomes along with hybrid error correction using the short, high-fidelity sequences.

  • Jabba is a hybrid method to correct long third generation reads by mapping them on a corrected de Bruijn graph that was constructed from second generation data. It uses a pseudo alignment approach with a seed-and-extend methodology, using maximal exact matches (MEMs) as seeds12. The tool is available here: https://github.com/biointec/jabba.
  • HALC is a high throughput algorithm for long read error correction. HALC aligns the long reads to short read contigs from the same species with a relatively low identity requirement and constructs a contig graph. This tool was applied on E. coliA. thaliana and Maylandia zebra data sets and has been showed to achieve up to 41 % higher throughput than other existing algorithms while maintaining comparable accuracy13. HALC can be downloaded here:  https://github.com/lanl001/halc.
  • The HYBRIDSPADES algorithm was developed for assembling short and long reads and benchmarked on several bacterial assembly projects. HYBRIDSPADES generated accurate assemblies (even in projects with relatively low coverage by long reads), thus reducing the overall cost of genome sequencing. This method was used to demonstrate the first complete circular chromosome assembly of a genome from single cells of Candidate Phylum TM6using SMRT reads14. The tool is publicly available on this page: http://bioinf.spbau.ru/en/spades.

Due to the constant development of new long read error correction tools, La et al. have recently published an open-source pipeline that evaluates the accuracy of these different algorithms15. LRCstats analyzed the accuracy of four hybrid correction methods for PacBio long reads over three data sets and can be downloaded here: https://github.com/cchauve/lrcstats.

Sović et al. evaluated the different non-hybrid and hybrid assembly methods for de novo assembly using nanopore reads16. They benchmarked five non-hybrid assembly pipelines and two hybrid assemblers that use nanopore sequencing data to scaffold Illumina assemblies. Their results showed that hybrid methods are highly dependent on the quality of NGS data, but much less on the quality and coverage of nanopore data and performed relatively well on lower nanopore coverages. The implementation of this DNA Assembly benchmark is available here: https://github.com/kkrizanovic/NanoMark.

References:

  1. Rhoads, A. & Au, K. F. PacBio Sequencing and Its Applications. Genomics, Proteomics Bioinforma. 13, 278–289 (2015).
  2. Koren, S. et al. Hybrid error correction and de novo assembly of single-molecule sequencing reads. Nat Biotech 30, 693–700 (2012).
  3. Bashir, A. et al. A hybrid approach for the automated finishing of bacterial genomes. Nat Biotechnol 30, (2012).
  4. Goodwin, S. et al. Oxford Nanopore sequencing, hybrid error correction, and de novo assembly of a eukaryotic genome. Genome Res 25, (2015).
  5. Au, K. F., Underwood, J. G., Lee, L. & Wong, W. H. Improving PacBio Long Read Accuracy by Short Read Alignment. PLoS One 7, e46679 (2012).
  6. Au, K. F. et al. Characterization of the human ESC transcriptome by hybrid sequencing. Proc. Natl. Acad. Sci. 110, E4821–E4830 (2013).
  7. Weirather, J. L. et al. Characterization of fusion genes and the significantly expressed fusion isoforms in breast cancer by hybrid sequencing. Nucleic Acids Res. 43, e116 (2015).
  8. Ning, G. et al. Hybrid sequencing and map finding (HySeMaFi): optional strategies for extensively deciphering gene splicing and expression in organisms without reference genome. 7, 43793 (2017).
  9. Steijger, T. et al. Assessment of transcript reconstruction methods for RNA-seq.(ANALYSIS OPEN)(Report). Nat. Methods 10, 1177 (2013).
  10. Tilgner, H., Grubert, F., Sharon, D. & Snyder, M. P. Defining a personal, allele-specific, and single-molecule long-read transcriptome. Proc. Natl. Acad. Sci. 111, 9869–9874 (2014).
  11. Beckmann, N. D., Karri, S., Fang, G. & Bashir, A. Detecting epigenetic motifs in low coverage and metagenomics settings. BMC Bioinformatics 15, S16 (2014).
  12. Miclotte, G. et al. Jabba: hybrid error correction for long sequencing reads. Algorithms Mol. Biol. 11, 10 (2016).
  13. Bao, E. & Lan, L. HALC: High throughput algorithm for long read error correction. BMC Bioinformatics 18, 204 (2017).
  14. Antipov, D., Korobeynikov, A., McLean, J. S. & Pevzner, P. A. hybridSPAdes: an algorithm for hybrid assembly of short and long reads. Bioinformatics 32, 1009–1015 (2016).
  15. La, S., Haghshenas, E. & Chauve, C. LRCstats, a tool for evaluating long reads correction methods. Bioinformatics (2017). doi:10.1093/bioinformatics/btx489
  16. Sović, I., Križanović, K., Skala, K. & Šikić, M. Evaluation of hybrid and non-hybrid methods for de novo assembly of nanopore reads . Bioinformatics 32, 2582–2589 (2016).

 

Choosing the Right NGS Instrument for Your Research

If you’re about to embark on a high throughput sequencing project, choosing the right sequencing instrument to use is an important consideration. Perhaps you’re replicating a published study or repeating an experiment from previous work and the instrument you plan to use is known. If not, the right sequencing instrument should be based on the sequencing goal you are trying to achieve. Instrument features to take into consideration include: number of reads per run, read length, read type (paired or single end), error type, turnaround time and price. Using Genohub’s Shop by Project page, you can enter the number of required reads or coverage you need and instantly compare instruments, filtering by read length and sorting by turnaround time and price. To get a better idea for the differences between NGS instruments, we’ve generated the following comparison: Table 1.   

Certain instruments are ideally suited to specific applications. Illumina instruments are versatile and ideal for a variety of sequencing applications, including: de novo assembly, resequencing, transcriptome, SNP detection and metagenomic studies. The HiSeq and GAIIx instruments are both suited for analyzing large animal or plant genomes. High level multiplexing of samples are possible when analyzing species with a smaller genome size. While the Illumina MiSeq outputs significantly fewer reads (Table 1), its read lengths are significantly longer making it ideal for small genomes, sequencing long variable domains or targeted regions within a genome. The only real limitation to the Illumina platform is its relatively short reads compared to other platforms (Roche 454 and PacBio).

The Ion PGM (Ion Torrent), is ideal for amplicons, small genomes or targeting of small regions within a genome. Its low throughput makes it ideal for smaller sized studies. The Ion Proton however is capable of generating significantly larger outputs (Table 1) making sequencing of transcriptome, exome and medium sized genomes possible.

The PacBio RS/RS II breaks the mold of other short reads high throughput sequencing instruments by focusing on length. The reads, averaging ~4.6 kb are significantly longer than other sequencing platforms making it ideal for sequencing small genomes such as bacteria or viruses. Other advantages include its ability to sequence regions of high G/C content and determine the status of modified bases (methylation, hydroxymethylation) without necessitating the need for chemical conversion during library preparation. The instrument’s low output of reads prevent it from being useful for assembly of medium to large genomes.

The Roche 454 FLX+ is typically used in studies where read length is critical. These include de novo assemblies of microbial genomes, BACs and plastids. It’s long read length has made it a favorite of those examining 16S variable regions and other targeted amplicon sequences. The lower output of the FLX and FLX+ instruments make it less cost-effective for transcriptome or larger genome studies. Roche has announced that it will stop producing the 454 in 2015 and end servicing in mid-2016. 

The SOLiD series of instruments are high throughput, generating a large number of short reads. De novo sequencing, differential transcript expression and resequencing are all viable applicaions of the SOLiD platform. The weakness of the platform is its short reads making assembly very difficult. 

If you’re still not sure about what NGS instrument to choose for your work, feel free to contact us for our complementary sequencing project consultation