Hybrid assembly with low coverage PacBio libraries

[dsher]

Hi everyone,

We are working on sequencing the genomes of a few bacteria and microalgae. We recently had a PacBio run which provided pretty bad results (only 76 MB of data - ~15x coverage of our ~4.5MB genome, average read length ~5,800 bp) which could not be assembled into anything coherent using HGAP. Canu assembler provided what looked like better results (106-110 contigs, N50 ~23kb) but these still covered less than one-half of the expected genome size (sum of contig length was ~2.3 MB, and our genome is estimated at ~4.5MB).

Given that we are having issues with getting high-quality DNA for PacBio from some of our cultures, we were thinking of making an Illumina library to use these reads for a hybrid assembly. Does anyone have any experience with using such a hybrid method with low-coverage PacBio data? Specifically, are there any recommended tools or parameters suggested for such an assembly?

One point to keep in mind is that our microalgae cultures are not axenic so we expect a (low) level of contaminating DNA.

Any recommendations will be very helpful.

Thanks
Daniel

[capricy]

Hi, Daniel,

I have similar question. My illumina reads assembly covered about 50% of the genome. Do you have some updated ideas to share?

Thanks.

C

[dsher]

Hi Capricy - no updated ideas yet... How do you know your Illumina assembly covers only 50% of the genome? What kind of libraries did you use, and what coverage did you get on the assembled parts?
Cheers
Daniel

[capricy]

The data were from E coli. New isolates. Library is illumina miseq. Coverage was estimated from the assembly/reference alignment.

[dsher]

I don't know much about the genomics of different E. coli strains but isn't that coverage relatively low? If you just mapped the reads to one or more of the known E. coli genomes, what coverage would you get? That might help you see whether you have a very weird genome (e.g. very large genomic islands which do not map to your reference sequence) or whether your assembly isn't very good. With our larger cyanobacterial genomes I think the low coverage is because the PacBio assembly isn't very good.

BTW we recently published a comparative analysis of what you get when you close a genome compared to fragmented de-novo assemblies for a different heterotrophic bacterium (http://journal.frontiersin.org/artic...016.00248/full). This has some information on the number of sequences and coverage we got that might be worth comparing with your results.