Good coverage value for de novo assembly with NanoPore Reads?

Question

[marianaer98]

Hello, I am trying to do a de-novo assembly of a turtle genome of about 2.15 gbases with Nanopore Minion reads. I have read about Canu as a good tool for this purpose, but it seems it yields good a good assembly with reads of at least 20x coverage, and the coverage of my reads is only about 2.5x. Is this value too low for de novo assembly with Nanopore reads? What would be a good coverage value?

Thank you everyone

Answer 1

[colindaven]

Yes, 2.5X is too low for de novo.

For structural variant detection vs a reference genome about 10X is recommended.

For de novo I would suggest 30-40X, whereby the longest reads are the most important.

Sorry, but 2.5 X is likely only maybe useful for somehow scaffolding Illumina contigs if you have them (difficult, poor results expected).

You can print the read distribution eg by using stats.sh from bbmap so we can have a look if you like.

The tool Miniasm or wtdbg2 https://github.com/ruanjue/wtdbg2 might give you a very very sketchy assembly but I doubt it.

[Picasa]

Hi,

For de novo I would suggest 30-40X, whereby the longest reads are the most important.

Do you have a source for this ? thanks.

[colindaven]

Pretty sure these figures come from work from the group around Adam Philippy who wrote Canu. Makes sense though, 2.5X is no way near enough to actually correct long reads from, and self corrected reads are needed before the assembly overlap step. They have published theoretical curves on effect of read coverage on NG50 , check the Canu and alternatively Flye and wtdgb2 papers.