HiFi reads

[rsandean]

Hi, I was wondering if there was any way of implementing HiFi reads only?

[OmarOakheart]

Hello,

If you mean PacBio HiFi reads, I've never worked with PacBio data before so I'm not familiar with how the data looks - but if it's a situation where HiFi reads are the highest-quality reads, you could just start by subsetting your input FastQ file to only include HiFi reads, and then use this HiFi-only subset of reads as the input to nPhase.

Hope that's helpful!

[rsandean]

Thank you for the advice! I was also wondering what the "reference" meant here, is it a close diploid relative or the draft assembly of the polyploid? (sorry if this was obvious, I couldn't quite confirm this from the paper). I essentially wanted to check the percentage of switch errors in my phased tetraploid genome. I had originally made the vcf file with my hifi pacbio reads aligned against the diploid close relative which I was going to put through the partial pipeline but I see now that it's not recommended?

Thank you!

[OmarOakheart]

If you have a very contiguous de novo assembly, I would probably try to use that file as the reference. But I think you're saying that you have an already phased de novo assembly for a tetraploid?

I think it also depends on what your species is, but the "reference" file simply has to be a haploid reference genome, and then nPhase will map all of the reads to this reference and try to separate them into individual haplotypes. So you want this reference to be as contiguous as possible, and you want it to be close enough to the sample such that you expect all of your reads to be mapped.

Do you know if your tetraploid has any aneuploidy? If you think there's no aneuploidy, then I remember comparing nPhase to flopp (https://github.com/bluenote-1577/flopp) and finding that flopp is very accurate and very fast. The main advantage of using nPhase is that flopp requires you to know how many haplotypes are present in your sample, while nPhase will figure it out for you and tell you if, for example, chr2 has three, four or five haplotypes total. It will also tell you if you have 4 copies of chr2 but only three distinct haplotypes, or if one part of a chromosome only shows a few distinct haplotypes.

I think it's possible that nPhase partial won't properly parse your VCF input. You can still try it, and if it runs fully, just be critical of the output. Using the full pipeline is the most certain way of knowing everything ran properly. Are you working with a large genome? nPhase is unfortunately very slow with large genomes, if you are I can give you advice on how to reduce the runtime while maintaining performance as much as possible