Scaffolding hifiasm assembly with CLR?

Question

[Phlupp]

Hi all!

For a plant genome assembly project, I have 55x CLR, 28x HiFi and 14x Illumina reads.

Which workflow would be the best to get the most of the data?

Hifiasm alone resulted in 417 contigs with L90: 58 and N50: 20Mb. 96.8% Busco score

The best scaffold result is 307 scaffolds with L90: 46, N50: 30Mb, gaps: 0.015%, 96.7% Busco score. I have canu correcttrimmed the CLR reads and (Racon) polished them with the Hifi reads before using them to scaffold (LRScaf) above mentioned hifiasm assembly. Then I racon and pilon polished the scaffolded assembly to achieve above mentioned best result.

Is this already the best approach? Or is it possible that I introduced errors to the hifiasm assembly by adding the CLR information? I know that Hi-C would improve the hifiasm assembly much more, but I got the task to test out everything with the provided data.

I would be very thankful for your advice and thoughts on this!

Answer 1

[lieven.sterck]

Sounds like a decent and acceptable approach to me, also the stats of the result look fine (I assume few very large scaffolds, Mb size, and a lot very small ones?, not very uncommon though) .

You could consider switching Hifiasm to flye for the assembly step, but should not expect miracle improvements by doing so

What is the expected genome size of your species btw?

[Phlupp]

Thank you for your reply lieven.sterck!

It is nice to know, that I am on the right track.

Yes, the scaffold length histogram shows that I have around 30 large ones (60Mb - 10Mb) and the rest are small (<10Mb).

The estimated genome size is 1.2Gb with 15 chromosomes.

Okay, I am running a CLR only flye assembly right now to test it. Or do you mean a flye assembly with the HiFi reads?

[lieven.sterck]

I think flye accepts all those inputs (or can be tricked into accepting them all ;) ) , so I would personally run an assembly with all possible data to start off with.

Flye has quite decent documentation and 'protocols' so go and have a look at it.

[Phlupp]

Ah interesting, thanks for this tip lieven.sterck!

After reading through some closed issues about hybrid assemblies and the faq, docs I am running this hybrid approach:

flye --pacbio-raw $HiFireads $CLRreads $ILreads1 $ILreads2
--iterations 0 
... 

Then resume flye-polishing with HiFireads as recommended in the faq (under hybrid assembly with HiFi and ONT).