Scientists at the University of Oslo’s Centre for Ecological and Evolutionary Synthesis (CEES) have used multi-kilobase sequence reads from the PacBio® RS sequencer to produce a dramatically improved genome assembly for the Atlantic cod. In many ways the cod genome seemed like a puzzle that might never be fully solved, but the Single Molecule, Real-Time (SMRT®) Sequencing platform made significant inroads — and just in time, as the team of researchers working on cod recently received funding to resequence 1,000 more of them. Being able to base these new efforts on a reliable genome assembly will make future results far more meaningful.
Lex Nederbragt, a research fellow at the University of Oslo and a member of the Norwegian High-Throughput Sequencing Centre, says that in the last decade or so, “there has been a growing interest in the genomics of this organism.” Cod is the most important aquatic species in Norway and other commercial fishery nations. Also, cod has an interesting population ecology; some populations do exceedingly well, whereas others get depleted through fishing and never recover to historic abundances. A good genome assembly can aid in finding those regions that influence traits important for disease resistance and growth rates, which may prove crucial for the economic success of the aquaculture industry.
In 2008, Nederbragt and his colleagues Bastiaan Star, Sissel Jentoft, Kjetill S Jakobsen, and others from the CEES-led Cod Genome Sequencing Consortium began a cod genome project using shotgun and mate-pair sequencing on the 454® platform. They mixed in some long-range information from BACs sequenced using traditional Sanger sequencing that resulted in an assembly having thousands of scaffolds and hundreds of thousands of contigs for the 830 Mb genome. Some 35 percent of the bases in the scaffolds were gaps, Nederbragt says, which of course proved quite a challenge for the Ensembl annotation team.
“They managed to produce a meaningful annotated genome by taking well-known genes from stickleback and other fishes to try to put together the missing pieces in cod,” he adds. In generating an assembly and annotation, the project was a success; but scientists knew that for certain regions of the genome, the genes would not accurately represent the cod genome.
Even while the first cod genome assembly was being published, Nederbragt and his colleagues were casting about for ways to improve it. One challenge was the marked heterozygosity of the wild-caught, diploid cod being sequenced. “Besides the SNPs that you would normally expect, we see large differences over hundreds of bases — sometimes even kilobases — either missing from the other chromosome, or causing differences in regions when we align them,” Nederbragt explains. “This confuses assembly programs.” The fish also had many short tandem repeats (STRs).
When the Oslo center acquired the PacBio RS in 2012, Nederbragt and his colleagues tested out the instrument by running their default cod sample. “When we looked at these PacBio reads mapping to the assembly, we saw them crossing large gaps of even multiple kilobases,” he says. It was a moment the team had been anticipating for years. “I could see that the problem of STRs and heterozygosity could be addressed by this technology,” Nederbragt adds.
Layering the PacBio and previous reads together, and using the highly accurate consensus, the team generated very long reads, error-corrected them using the short read data, and ran them through Celera® Assembler. “We’ve never seen a faster assembly,” Nederbragt says; it came together in just 36 hours.
As Nederbragt and his colleagues sifted through the new assembly, they realized that the assembler was splitting haplotypes rather than merging them, so the heterozygous regions were being run as linear sections of the genome, rather than alternates of the same section. “The sequencing problem is now gone; it looks like we have the whole genome present in PacBio reads,” Nederbragt says. “Now it has become a bioinformatics challenge.” He and his team are currently working to quantify the regions they believe should be split into haplotypes and to figure out the differences between them.
As they determine the best bioinformatic solution to the assembly, they are starting to investigate the new genome data to see where it varies from the original stickleback-oriented assembly. They’ve already seen an exon in the original annotation that potentially does not exist in the all-cod assembly, Nederbragt says, noting that a full comparison of the two genome assemblies will take place in the future. For now, they are focused on getting this new assembly into its 23 pseudochromosomes, which can then be shipped off for annotation. “The goal is to get the annotators an assembly good enough that they don’t need to retrofit it with information from other organisms,” he says.
For more on this work, see the full case study.