Low level of divergence in the sample being assembled

[YuanwenGuo]

Hello,
I am trying to assemble ~20 linear plasmids which are very similar to each other, the whole plasmids (average length ~16kbp) are almost identical to each other, except for a single region (~20bp) within each plasmid is different. In other words, all the plasmids have same backbone, just the insert part is different. I already tried to reduce correctedErrorRate to 0.13 (we have OXN data), but we still can not get all 20 sequence in the final assembly, we have only ~10.

I realized this is a special case for any genome assemblers, but since we are having really awesome results with Canu for regular genome assembly, so we would like to give it a try.

I am just wondering how Canu will treat this localized 20 bp mismatches during assembly, and is there any parameters I can adjust in Canu to make it work? My current commands are:
canu maxMemory=80 redMemory=50 oeaMemory=50 gridOptions="--time=100:00:00 --partition=**" -p $prefix -d $dir genomeSize=320k correctedErrorRate=0.13 gnuplotTested=true -nanopore-raw $OXN_file

Just in case it might help, we have high coverage raw data, about a few thousands X.

Thank you!
Yuanwen

[skoren]

A 20bp difference out of 16kb is quite small given nanopore error rates, in general I'd expect the correction to corrupt that difference. Ideally, I'd suggest something similar to what I suggested in #1885. That is, use a consensus plasmid to map reads + call variants. The variant callers can give you reads supporting each variant so you can bin + assemble subsets rather than mixing them all.

You could also try re-calling the ONT data with a newer basecaller (like bonito or recent guppy versions) which can be assembled w/o correction due to their higher accuracy as I suggested in #1715 (-untrimmed 'batOptions=-eg 0.12 -sb 0.01' 'correctedErrorRate=0.12' 'maxInputCoverage=100' -pacbio-hifi <your nanopore fastq>) assuming you're using Canu 2.1.1

[YuanwenGuo]

I appreciate the suggestion! We probably will try the targeted assembly strategy first to see how it works.

Best,
Yuanwen