Identifying AxB F1 hybrids in the D8 2018 individuals
I was interested in identifying putative AxB F1 hybrids in the 2018 D8 individuals. First I took all A and B individuals from D8, regardless of year. I then found all SNPs that were fixed reference in A and fixed alternate in B, or vice versa. I next pulled out all the 2018 D8 individuals and called those same SNPs, and polarized the genotypes as B (dos0), A (dos2), or heterozygous. Not surprisingly, all 7 D8 2018 A individuals were 100% A. Another 8 individuals had a prop het between 0 and 0.9. These could be recombinant hybrids between A and B, or something else. March20_2018_D8_16 and March20_2018_D8_33 look a bit like backcrosses for example. March20_2018_D8_37 could be a F1 hybrid with high error rates in calling heterozygotes.
Another 43 individuals look like they could be F1 hybrids between A and B. They are over 90% heterozygous for SNPs fixed between A and B. This is more F1 hybrids than you had on your piecharts Alan, why the discrepancy?
I was worried about all the homozygous calls, but I was thinking those could be due to low read depth sites and so calling a heterozygous site as homozygous in error. To look at this, I choose a few representative individuals and looked at the distribution of read depth at individual variants depending upon whether they were called as A, B, or heterozygous. We see that overall A and B sites do have lower read depth. So this makes me feel pretty confident in saying all these individuals are AxB F1 hybrids. Interesting that in March 2018 you get both F1 hybrids, As, and backcross/other looking things. But later in 2018 you get only F1 hybrids (though sample size is lower). But in December 2018 (after the pond likely went dry), the single individual is recombinant.
I also wanted to check the distribution of read depths at A, B, and heterozygous sites in individuals that don't look like F1 hybrids. Here for example, is one that looks more like a F1 (March20_2018_D8_18). Looks like you would expect.
Ok, so all this taken together. Do you agree that the 43 individuals in the original screenshot look like F1 hybrids? Are there reasons to think that they are not? Is further testing necessary?
For my next steps I am thinking about taking single representatives of A, and B, as well as a subsetted version of the 43 individuals above (one representative per superclone), and running those through the kinship program to see if we get the expected parent/offspring/sibling relationships back out. Agreed? Thoughts?
Thanks!
Another 43 individuals look like they could be F1 hybrids between A and B. They are over 90% heterozygous for SNPs fixed between A and B. This is more F1 hybrids than you had on your piecharts Alan, why the discrepancy?
I was worried about all the homozygous calls, but I was thinking those could be due to low read depth sites and so calling a heterozygous site as homozygous in error. To look at this, I choose a few representative individuals and looked at the distribution of read depth at individual variants depending upon whether they were called as A, B, or heterozygous. We see that overall A and B sites do have lower read depth. So this makes me feel pretty confident in saying all these individuals are AxB F1 hybrids. Interesting that in March 2018 you get both F1 hybrids, As, and backcross/other looking things. But later in 2018 you get only F1 hybrids (though sample size is lower). But in December 2018 (after the pond likely went dry), the single individual is recombinant.
I also wanted to check the distribution of read depths at A, B, and heterozygous sites in individuals that don't look like F1 hybrids. Here for example, is one that looks more like a F1 (March20_2018_D8_18). Looks like you would expect.
Here is one for an individual that looks a bit more like an A backcross (March20_2018_D8_16). Again looks like you would expect.
Finally one that looks more like a B backcross (March20_2018_D8_40). Again looks like you would expect.
For my next steps I am thinking about taking single representatives of A, and B, as well as a subsetted version of the 43 individuals above (one representative per superclone), and running those through the kinship program to see if we get the expected parent/offspring/sibling relationships back out. Agreed? Thoughts?
Thanks!
Comments
Post a Comment