should I use longer --k1 and --k2 in where I want to correct ONT with HiFi?

[dukecomeback]

Hi, like told in the title. I am trying to correct ONT with HiFi. Do you think this should work? Should I use longer --k1 and --k2? Is there any other parameters I should be aware?

Thank you and best regards!

 Kang

[GuillaumeHolley]

Hi @dukecomeback,

Thanks for reaching out. This is an interesting experiment although I have no idea how Ratatosk will behave here. I would expect the indexing step to take longer than with short reads due to the increase of read length.

I would probably not change the size of k-mers --k1 and --k2 because even though you could have a more contiguous graph with longer kmers, something you can afford with your PacBio HiFi reads, Ratatosk still needs to anchor the ONT reads onto the graph using those k-mers and that's not something the error rate of ONT R9 reads would allow.

However, something I would definitely try would be to increase the insert size (-i, --insert-sz) to 5000, maybe even 10000, given that your PacBio reads are most likely at least that long.

Let me know how it worked.
Guillaume

[dukecomeback]

Hi Guillaume,

Thank you for your suggestions!

I have been trying this but have not got any results, cause with hifi reads, rotatosk need a huge of memory. So my jobs were always OOM killed by slurm. I am not sure if this is something related to hifi reads. I think I may have to try with short reads first to get an idea of rotatosk's normal behavior.

Thank you!
Kang