Higher QV in a_ctg than p_ctg

[pcarbone]

Thank you for the nice and very efficient assembler.

I assembled a highly heterozygous genome and hifiasm did very well separating primary and haplotigs (p_ctg size is only a bit larger than the expected genome size and a_ctg length is 90% of p_ctg size).

I ran Merqury for QC from short-read kmers of the same sample. I always have higher consensus accuracy QV in a_ctg (QV=~49) than p_ctg (QV=~37) independently of the parameters I use in hifiasm.
Is it expected?
Are there some hifiasm options that could improve QV of p_ctg?

Thanks

[chhylp123]

The QV of p_ctg should be roughly as same as that of a_ctg, and QV=~37 is too low for HiFi assembly. Could you please evaluate QV with yak qv (https://github.com/lh3/yak)? I'm not sure if it is the problem of hifiasm or Merqury.