We have a hybrid illumina + #nanopore assembly of a culture that resolved into 2 contigs. We're looking for the best way to close the genome. More long read (the DNA was more fragmented than ideal)? Or develop primers at the contig ends and do PCR? Other options? - ThreadSky

uncultured.carinilab.com • 45 days ago

We have a hybrid illumina + #nanopore assembly of a culture that resolved into 2 contigs.

We're looking for the best way to close the genome. More long read (the DNA was more fragmented than ideal)? Or develop primers at the contig ends and do PCR? Other options?

Comments

born2bewildtype.bsky.social•44 days ago

Raven is more relaxed in creating overlaps between contigs than unicycler and Flye in my experience if those dont work. Remember to manuel inspect the ends for hints to why they dont resolve into a single contig.

bioinformer.bsky.social•36 days ago

Paul - we’ve done hybrid whole genome sequencing for about 4,000 different strains at this point. One thing we’ve learned is sometimes you just have manually scaffold contigs using different assemblies from the same data put through different pipelines. It’s time consuming, but worth the effort.

bioinformer.bsky.social•36 days ago

Feel free to DM me if you need additional help. Happy to set up a call and add in one of my team members who works on the ATCC Genome Portal assemblies.

bioinformer.bsky.social•36 days ago

Plus UofA is my Alma mater so… 🙌🫣 happy to help.

uncultured.carinilab.com•36 days ago

We'll give it a bit more time and reach out if we could use additional help. Thank you Jonathan!

bioinformer.bsky.social•36 days ago

Sounds good. Anytime Paul.

dnakendra.bsky.social•45 days ago

Hybacter is pretty easy to install/run if you haven't tried it yet.

uncultured.carinilab.com•45 days ago

I'll check this out! Thanks Kendra. Hope spring is treating you well!

blasarre.bsky.social•45 days ago

I would try additional assemblers. I work on highly complex genomes and different assemblers yield wildly different results from the same ONT read set. NextDenovo has worked well for me on chromosomes (though isn’t great for plasmids in my experience). Autocycler may be of help.

uncultured.carinilab.com•45 days ago

This is a unicycler assembly. We'll try some other tools. High GC content genome too 70%. So this might do the trick! Thank you!

germusthermophilus.bsky.social•45 days ago

Tricycler should do much, much better!

paulmorwin.bsky.social•44 days ago

You’ve probably seen this but I have had some good success following this https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1010905

gringene.org•45 days ago

I agree with the other replies saying that there's no need to do more sequencing.

If you've got it down to two contigs, the final stitching bits become simple enough that they can be done via in-silico experimentation.

carmcdhume.bsky.social•43 days ago

Just tagging @tom-mathers.bsky.social in case he's able to advise!

mtgassr.bsky.social•45 days ago

Any idea what you’re dealing with? I’ve dealt with near identical duplication of a prophage. When most of it is resolved, I’ve been able to manually close by looking at ONT reads that map to just the contig ends. If not, then had to redo ONT to get read lengths that spanned the entire region ~40kb

uncultured.carinilab.com•45 days ago

70% GC bacterial genome. We know "who" it is but prefer to not say much on that yet. I'll try looking at the ends and see if I can find some overlap.

mtgassr.bsky.social•45 days ago

Fun! That’ll do the trick a lot of the times with reads in to the 1-10Kb range if you don’t think you can easily get longer reads. I’ll align reads to last 500bp of contig ends and assemble those reads on their own. Then visual those contigs with the originals in Mummer to see if they’re bridging

gutmichaelbio.me•37 days ago

Maybe @stevenjrobbins.bsky.social has some ideas?

stevenjrobbins.bsky.social•36 days ago

Like the others have said, trying a new assembler can work wonders. Have seen MetaFlye assemblies turn out like hairballs (think that's what Unicycler uses?) and Hifiasm spits out a complete genome. Doesn't always happen.

Additional long read data can also help.

stevenjrobbins.bsky.social•36 days ago

Not universal that long read-only assemblies are better than hybrid though. The long read-only assemblers often fail to recover plasmids, so if you expect them, that's a consideration.

kirk3gaard.bsky.social•36 days ago

Just remember to go with the rapid preps and your plasmids will be fine? Ligation based chemistry requires ends.

stevenjrobbins.bsky.social•36 days ago

Collaborator believe did use rapid, did not get the plasmid using all assemblers. Tested a few.

uncultured.carinilab.com•36 days ago

Thanks Steven! I think we try a bit more long read data next. We have nano & illumina now. We have easy access to PacBio seq here at UofA. The DNA going in to sequencing was a bit more sheared and a lower amount than ideal. We've tried several assemblers & settings and get 2-100 contigs.

stevenjrobbins.bsky.social•36 days ago

Sounds like the way to go if you’ve already tried a few assemblers. That’s one nice thing about PacBio, their size selection is harsh by necessity, so short fragments not a huge issue. ONT will preferentially sequence the small stuff if you don’t remove it.

titus.idyll.org•36 days ago

have you tried using bandage to look at the assembly graph? there may be a way to manually resolve things. Not my usual go-to but I know Jill Banfield does this kind of thing a lot :)

uncultured.carinilab.com•36 days ago

Yes, I've put them into bandage, but need to read up on how to use it to do what you suggest, because it's not obvious to me. End of semester shenanigans are ending this week, so 🤞 I'll have time to play around. Thanks Titus!

titus.idyll.org•36 days ago

great!

weisbergaj.bsky.social•45 days ago

Have you tried different assembly methods? Sometimes we get better assemblies doing nanopore assembly first (flye/autocycler) then polish with Illumina reads, sometimes we get better with a short/long hybrid approach (unicycler).

weisbergaj.bsky.social•45 days ago

You can also check assembly graphs with Bandage to see if the assembly is complete and there are plasmids, or alternatively if those contigs go together. PCR probably would be tricky since the repeats are likely longer than you can amplify. I would try SPRI size selection and resequencing the DNA.

uncultured.carinilab.com•45 days ago

We assembled with Unicycler. Resolved to two incomplete contigs of the same GC content (which was anomalously high at 70%). 3.2 Mbp & 1.3 Mbp. Looking at the gene annotations, it appears that they are the same organism. Haven't ruled out 2 chromosomes, but no reason to think that yet.

weisbergaj.bsky.social•45 days ago

I would try Flye with the nanopore reads and see if that puts them together. Flye tends to be pretty good even with low coverage. Unicycler relies on the SPAdes assembly of Illumina reads first, then scaffolds with long reads. That works if the Illumina assembly is good but not always.

surtlab.bsky.social•45 days ago

I would try bottlenecking down to a single colony and reisolating DNA too…weird repeat stuff happens and sometimes you gotta get lucky with the culture and extraction

surtlab.bsky.social•45 days ago

(Don’t Illumina again, obvs)

uncultured.carinilab.com•45 days ago

They don't grow on plates (yet?). Broth low-ish density enrichments. Barely a pellet.

geomicrosoares.bsky.social•37 days ago

How deep is the ONT data? If R10.4.1 you’ll probably be better off just assembling the long reads these days tbh. If oversequenced consider sub sampling as @rrwick.bsky.social recommends in his tutorial https://rrwick.github.io/2023/11/06/accuracy-vs-depth-update.html

uncultured.carinilab.com•37 days ago

It's not oversequenced—we can improve on both depth and fragment size. The input amount was quite low. I think we'll try resequencing with higher quality DNA if we can get it, reassemble and re-evaluate.

cyan0bacteria.bsky.social•44 days ago

Have you tried a generative AI model? I hear they will replace all experiments within 6 months 😂.

Comments

Posting Rules

Reply