check resulting assembly from haplotype binning is better than without

[chhylp123]

check resulting assembly from haplotype binning is better than without

[zhenzhenyang-psu]

I compared the N50 of the fast from the following three files:
comb.hifiasm.p_ctg.fa
haplotype_binning_-3-4out.hap1.p_ctg.fa
haplotype_binning_-3-4out.hap2.p_ctg.fa

The N50 is not much different:
combined primary contain N50:
"totalContigLength": "1419992",
"numberOfContigs": "3",
"contigN50": "492778",
"longestContig": "657757",
"totalScaffoldLength": "1419992",
"numberOfScaffolds": "3",
"scaffoldN50": "492778",
"longestScaffold": "657757",

haplotype1 primary contain N50:
"totalContigLength": "1419063",
"numberOfContigs": "3",
"contigN50": "491528",
"longestContig": "658078",
"totalScaffoldLength": "1419063",
"numberOfScaffolds": "3",
"scaffoldN50": "491528",
"longestScaffold": "658078",

haplotype2 primary contig N50:
"totalContigLength": "1416418",
"numberOfContigs": "3",
"contigN50": "492613",
"longestContig": "655973",
"totalScaffoldLength": "1416418",
"numberOfScaffolds": "3",
"scaffoldN50": "492613",
"longestScaffold": "655973",

They are not much different. I guess this means that the haplotype binning information I provided is not better than what hifiasm can do inherently.

so presumably if the HIFI reads that I provide span a much longer haplotype block than what HIFI reads can phase, it should be better.

Anyways, sorry for bothering.

I was wondering the difference of primary unitigs (p_utg.gfa)/the alternate unitigs (r_utigs.gfa) from primary contigs.
is there a way to visualize the .gfa for these different files for illustration.

Finally, there is a question of how to use the overlap information
(comb.hifiasm.ovlp.source.bin/comb.hifiasm.ovlp.reverse.bin) to identify any HiFi reads that overlap with HiFI reads of interest (among the input HiFi reads)?
thanks a lot,
zhenzhen

[chhylp123]

The assemblies with/without -3 and -4 are totally different:

1. If you run hifiasm in default, it will output a primary assembly including haplotype switch between two haplotypes. So the primary assembly is not a fully phased assembly, it is a mixture of two haplotypes.
2. If you run hifiasm with -3 and -4, it will try to output two fully phased assemblies, one for each haplotypes (see Figure 1 and Figure 2 in https://arxiv.org/pdf/2008.01237.pdf).

If the N50s of hap1, hap2 and the primary assembly are similar, I guess the reason might be that your sample is too simple. You can have a try with large and complex samples.

As for the overlaps, hifiasm can output all overlaps in PAF format with the option --write-paf.

[zhenzhenyang-psu]

thanks, Haoyu!
When you say large and complex samples, do you mean input file containing HiFi reads across a bigger region from the genome?

[lh3]

Option -1/-2 or -3/-4 is intended for whole-genome phasing.

[zhenzhenyang-psu]

I see. Thanks!
I wonder when using -3/-4, if the ability to designate reads from two separate haplotypes is limited - i.e. only a portion of the reads are indicated to be haplotype-specific and provided via -3/-4, how does hifiasm handle the remaining reads? Does it consider all the remaining reads as homozygous and add to each of the two haplotypes?

[chhylp123]

No, in this case, it can still check which read is heterozygous, and then assign heterozygous reads to one haplotype randomly.

[zhenzhenyang-psu]

No, in this case, it can still check which read is heterozygous, and then assign heterozygous reads to one haplotype randomly.

that's wonderful! thank you!