Strange BUSCO duplicate in hifiasm -l3 only vs hifiasm -l1+purge_dups about hifiasm HOT 3 OPEN

I feel like busco results might be not such accurate. In most cases, duplicated genes come from the corresponding homologous regions which haven't been fully purged. Probably you can check how do the corresponding nodes of these duplicated gene looks like in the assembly graph. I guess you should have a sense if these genes come from the homologous regions.

from hifiasm.

Taking a look at the GFA in Bandage, the specific contigs do not share edges (no bubbles or forks). None of the contigs have unusual GC content or read average read depth either. I am currently running a mummer alignment to gauge the level of sequence level similarity.

Interestingly, scaffolding the purge_dups assembly with 5 rounds of ntLink using the hifireads takes this BUSCO from being single copy complete to having 5 duplicates, 2 on the positive and 3 on the negative strands across 4 scaffolds. So using hifireads to gap fill and scaffold seems to reintroduce the duplicates. Lowering the s parameter to 0.35 in hifiasm does not lower the number of repeats.

from hifiasm.

Thanks... So it is hard to say

from hifiasm.