Is the WGDI based sub-genome phasing really reliable?

[GurjotSinghSidhu]

I was testing the WGDI based sub-genome phasing on an already phased and published assembly from an allotetraploid finger millet (Eleusine coracana) https://www.nature.com/articles/s41467-023-38915-6. I am using a closely related diploid, Oropetium thomaeum as an outgroup. I followed the manual according to the instructions given and after running the karyotype mapping (-km) function, I obtained the ancestor.txt file for finger millet, as shown here:

As per the instructions (to label the homoeologs with different numbers), I labelled all the chromosomes belonging to sub-genome A as 1 and all the chromosomes belonging to sub-genome B as 2.

After this I ran the polyploidy classification (-pc) and alignment (-a) steps, and got the sub-genome classification. All the chromosomes belonging to sub-genome A were colored red while all the chromosomes belonging to sub-genome B were colored blue, which led me to believe that the analysis phased the sub-genome accurately.

But then just as a test, I changed the labelling in the ancestor.txt file for finger millet - for the first four and last one (9th) chromosomes, I inverted the order (i.e. labelled sub-genome A as 2 and sub-genome B as 1), while for the chromosomes 5, 6, 7, and 8, kept the same labelling (i.e. sub-genome A as 1 and sub-genome B as 2).

And to my surprise, after running -pc and -a steps, I got the results where the sub-genome was phased totally incorrectly.

I tried it on several other genomes I am working on, and I found that what-ever chromosome is labelled as 1 in the ancestor.txt file is colored red and whatever is labelled as 2 is colored blue, while this method (WGDI based sub-genome phasing) claims that the red and blue should corrrespond to phased sub-genomes. Now, I am doubting the reliability of this pipeline. Any comments would be helpful.

[zhangrengang]

No need to be surprised; this behavior is expected and points to the core mechanism of the $\text{WGDI}$ pipeline.

• When you edit the $\text{ancestor.txt}$ file, actually you are manually assigning the subgenomes.
• The $\text{WGDI}$ pipeline does not automatically assign subgenomes (i.e., it doesn't calculate which chromosome belongs to "A" or "B").
• The $\text{-pc}$ and $\text{-a}$ steps just serve as visualization tools for your manual input; they do not perform the $\text{de novo}$ assignment of subgenomes. They simply color the chromosomes: 1 = Red and 2 = Blue, based on your provided $\text{ancestor.txt}$ file.
• You must rely on other evidence (which can be generated using $\text{WGDI}$, such as similarity with a diploid progenitor) to guide your manual assignment in $\text{ancestor.txt}$ before running the $\text{-pc}$ and $\text{-a}$ steps. You can see these rationales in this repo.

For your neo-allotetraploids, the better choice is the $\text{SubPhaser}$ pipeline in this repo, which is automatic and results in absolute phasing. It is designed to identify subgenomes and inter-subgenome exchanges algorithmically, but does not work for paleo-allopolyploids.

[zhangrengang]

The reliability of the $\text{WGDI}$ pipeline for subgenome phasing is directly contingent upon whether you have reliable, external evidence to accurately edit the $\text{ancestor.txt}$ file, which is the step where you are assigning the subgenomes.

[GurjotSinghSidhu]

Thanks for the answer. Can you please elaborate on the process of generating the ancestor.txt file - I tried to follow the example given on WGDI page but that seems difficult to follow. For practice, I am trying to generate the ancestor.txt file for the maize genome. Here is the comparison I have done with the Sorghum (I have filtered the matches to just Sb:Zm::1:2 using OrthoIndex filtering). Your help would be appreciated.

[zhangrengang]

The first ancestor.txt file for the reference or reconstructed ancestal genome (e.g., Sb) can be generated by manual editing. Then, the ancestor.txt file for other genomes (e.g., Zm) can be generated by mapping to the reference/ancestal genome using wgdi -km.

PS. the labels, Sb and Zm, should be reversed on the first dot plot.

[zhangrengang]

After mapping with wgdi -km, you still need to edit the subgenome column in the ancestor.txt file for other genomes (e.g., Zm).

[GurjotSinghSidhu]

Thanks for pointing out the label error, I have fixed it:
 

From the above dot plot, I can come up with the following analysis, i.e. how the 20 ancestral chromosomes (Sorghum is being considered to be a representative of the ancestral genome here) rearranged to produce the 10 chromosomes in extant maize. In the table below, 20 ancestral chromosomes are represented on the left and 10 extant maize chromosomes are represented on the top. If an ancestral chromosome contributed to a maize chromosomes, it is indicated with '1' otherwise 'empty cell'. The totals at the ends represent that from how many ancestral chromosomes each extant Zm chr. got contributions from OR each ancestral chromosome contributed to how many extant Zm chromosomes.

 

My goal is to create the ancestral karyotype with 20 chromosomes (based on maize sequence) and then determine how it re-arranged (fissions/fusions) to produce the extant maize genome with 10 chromosomes. In my opinion (I might be wrong), WGDI documentation is not super-helpful and that's why I am stuck with how to proceed to achieve my goal. Your help would be appreciated. Thanks!

[zhangrengang]

For your goal, you need to reconstruct the ancestral karyotype, and then trace the evolutionary trajectory based on the EEJ-NCF-RT model (but not the fissions/fusions model). You can see the methods in my paper or recent papers of Dr. Pengchuan Sun.