Separation of haplotypes in genome assembly

Question

[sansan96]

Hello everyone,

I have a genome from a highly heterozygous diploid plant that was assembled using pacbio reads with Canu in which haplotype differences were ignored during the assembly process. I would like to know if it is possible to separate this genome into haplotypes and what tools I could use. I will be very grateful for any guidance you can give me.

Answer 1

[Michael]

I don't think haplotypes can be reconstructed from collapsed assemblies without the raw data (if that is what you are asking), as the collapsed may not contain information on heterozygous variants needed to reconstruct haplotypes. If the raw data are available, why not generate a new haplotype-resolved assembly from the raw data, e.g. with HiFiassm or Flye -> HapDup?

An alternative could be to phase all reads into haplotypes first using nPhase and then assemble each cluster separately using haploid assembler.

High heterozygosity is a plus in this case, because it allows for better coverage of multiple heterozygous sites.

[sansan96]

Thank you very much for your answer,

I have the raw data and I am thinking of assembling with hifiasm, I only have a conceptual or methodological doubt. Hifiasm works with HIFI reads, the pacbio reads I have were generated in 61 SMRT sequencing cells PacBio, could I use hifiasm?

[gconcepcion]

hifiasm will not work for CLR data.

Your best option here is to run FALCON + FALCON_unzip to phase the data. This software pipeline has been unsupported for several years now so you will be on your own and your mileage may vary.

Depending on the genome size of your organism, this may or may not be feasible with available resources.

https://pb-falcon.readthedocs.io/en/latest/

Answer 2

[colindaven]

I haven't tried this, but there is software for this from the authors of Flye.

https://github.com/KolmogorovLab/hapdup

You'll need the (recent) long reads, and, as always, some element of luck, patience and decent hardware.