Assembling raw reads with Canu, but without correction

[ekarais]

Hello friends,

I have raw ONP and PacBio reads for the human genome, ecoli, and yeast, and I want to test Canu's performance on them, without correction.

If I understand correctly, I would only execute the assemble step of Canu, and specify -nanopore-raw or -pacbio-raw, depending on the technology.

Is this correct?

Also, what other parameters should I set/tweak to get better accuracy in this scenario? Some candidate parameters I know:

• rawErrorRate
• utgOvlErrorRate
• utgGraphDeviation
• utgRepeatDeviation
• utgRepeatConfusedBP

Also, is there a platform where I can ask this to the developers of Canu? Is github/issues the right place for such a question?

Any response greatly appreciated!

[colindaven]

Why not just specify -nanopore-corrected to probably make canu skip this stage ?

Also, why would you even want to do this ? You'll get really awful results, the correction is there for a reason.

Also, I think miniasm does rapid assembly on uncorrected reads, but then you have massive problems trying to correct base level errors in the asm afterwards.

[SNPsaurus]

wtdbg2 will overlap long reads and then correct/consensus as a separate step (I know this doesn't help getting Canu to do it that way, just following up on colindaven's comment).

Canu corrects reads before overlapping. I don't know if it can be shoe-horned into working on raw reads since it isn't expecting a 15% error rate when it overlaps. Just to offer additional annoyingly non-Canu advice, the flye assembler constructs an assembly graph before read correction.