# Canu assembly not making a single consensus?

I've downloaded reads from this BioProject. Using canu with default parameters (no correction), I've got 4 contigs, none of which really look like the reference plasmid here.

The command I used was:

canu -p ip40a -d ip40a_assembly -useGrid=false -maxThreads=6 -genomeSize=175k -nanopore-raw reads.fastq


Granted, the original publication for this sequence used minimap/miniasm but my questions are:

A. Why do I get multiple contigs? Doesn't canu generate a single assembled sequence?

B. This is a plasmid but canu says suggestedCircular=false. What's going on here?

Looking that the link you provided only 30% of the reads appear to be e. coli. I would suggest filtering the reads for those that align to e. coli and see where that gets you. Did the original publication describe the method they used for data cleaning?

A:

https://trace.ncbi.nlm.nih.gov/Traces/sra/sra.cgi?run=SRR3385327

cellular organisms: 30%
Bacteria: 30%
Proteobacteria: 28.09%
Gammaproteobacteria: 19.7%
Enterobacterales: 13.3%
Enterobacteriaceae: 12.54%
Vibrionales: 0.01%
Viruses: 1.17%


B:

Without a good assembly canu won't be able to determine if the assembled genome is circular. I would ignore this until you have a good example.

• Sadly the original publication only seems to say "assembled by miniasm/minimap" (along those lines) and I'm trying to figure out the pipeline as we speak -- there's no indication on whether the reads were error corrected, etc. Dec 2 '17 at 22:16
• Also, how would I filter the E. coli-identified reads? Dec 2 '17 at 22:26
• I would align the reads to the plasmid reference and see how many align and what the coverage of the reference looks like. Are there any regions of the reference that are uncovered. how even is the coverage across the reference. These are things that can help you understand what to expect from the assembly. Dec 3 '17 at 2:24
• It seems like it's actually the opposite that @thestatnoob needs to do: exclude E. coli reads, instead of selecting for them
– gringer
Dec 3 '17 at 5:40
• Correct he has the plasmid reference he is trying to assemble. He needs to select for those reads assuming the alignment coverage shows an assembly is even worthwhile. Dec 3 '17 at 6:35

The E. coli genome is closer to 4.6 Mb in size; if there are a large quantity of host reads in there, then they'll take over the reads and there won't be enough for your target plasmid. By default, it will only try to assemble from a maximum of ~40X coverage of the specified genome size (higher values do not add any more information for assembly).

It's possible that miniasm will take all reads into account, rather than just a portion of reads. You may need to filter out the E. coli reads for this assembly to work properly.

For what it's worth, Canu's default parameters do include a correction step.

• Yep, but the reads aren't for an E. coli genome. It's for a plasmid within E. coli, which is definitely within the 170-180k region (in fact, I got the number from the ENA...). I'm basically just trying to reconstruct the consensus plasmid that the authors have, using canu instead of miniasm. Dec 3 '17 at 0:46