Resolving Oversequencing with Flye

[gillianpatton]

Hi! I am currently using ONT long read to identify plasmids in Escherichia coli clinical isolates. I am getting decent coverage of my genomes (50-500x coverage) and am having minimal assembly problems. However, other people in my lab sequencing C. diff clinical isolates have been able to recover assemblies that failed assembly with flye by subsampling their ONT reads. They are currently subsampling to 50% of the reads present using filtlong, but we are interested in doing this filtering with the --asm-coverage flag in flye instead to help preserve our plasmids.

Any advice on whether oversequencing is a problem with long reads, and whether anyone has used the --asm-coverage flag to resolve oversequencing? We are also interested in achieving optimal basecalling accuracy for possible SNP analysis in the future. We are also polishing with short reads with polypolish.

Also, the --asm-coverage flag requires the use of the --genome-size flag as well. What happens if we slightly over or under-estimate the genome size? What exactly does flye use this flag to do on the back end?

[mikolmogorov]

Sorry for my late respnse! I don't think it should be a problem with assembly quality, the only disadvantage of very high coverage is longer running times. If you could give more details about the datasets that are having issues and provide log files, I may be able to give more advice.

[gillianpatton]

flye_barcode4_pass.log
flye_barcode4_fail.log

I have attached two flye logs for the same E. coli isolate. This isolate was run through flye twice, and it was only assembled when the parameters --asm-coverage 50 and --genome-size 5m were added. All other parameters were the same. We also had this issue for 6 other isolates.

We are also having an issue where some of our samples assemble, but their chromosomes are much too small to be E. coli (2.5-3.8 Mb instead of 5 Mb). It has also been suggested that subsampling our reads with filtlong may solve this issue, but we have not tried it.

[mikolmogorov]

I see that you are assembling the failed one with --nano-raw and pass with --nano-hq. This might be where the difference is coming from, although I don't see any obivious reasons for --nano-raw run to fail.

[gillianpatton]

Yes, that was a mistake. --nano-hq is the correct flag for our ONT chemistry (PromethION, R.10.4.1). We're exploring over-sequencing as a factor.