generalise `simulateAlleleFreq()` function to any ploidy about qtlseqr

[bmansfeld]

Hi Hugo,
Thanks for the detailed posts. It's been fun to collaborate and think on this with you.
Regardless of the ploidy levels. The rbinom solution is certainly faster and I used it in the SNP index simulations. I guess it eluded me for some reason in the allele freq sim. The way I went was kind of derived from the original slower code in the Tagaki et al scripts, which randomly sampled from a uniform dist and then asked if it was higher or lower than 0.5 to decide the allele.
I will probably incorporate the rbinom solution alongside the changes suggested in #5, just for the sake of speed.
In regard to polyploidy, and though I am not a polyploid expert the way you have it set up makes sense to me as is. However, I am aware that there are cases in autopolyploids that can behave in different ways such as multivalent pairing etc.
I will be in touch with Dr. Pat Edger in our dept who is our resident polyploidy expert in the coming weeks and see if there is a way to integrate a more accurate representation of the allele frequencies.
Maybe this is overkill and the method you suggested is perfect for the kinds of studies QTLseqr is for, I am just not confident enough about the polyploidy mess...
Let me know your thoughts.
Ben

[tavareshugo]

Hi Ben,

I'm no polyploidy expert either, so it seems sensible to discuss with someone more familiar with polyploid genetics! I suppose the way I put it made two assumptions:

• parents are fully homozygous
• segregation of the n copies in the F2 is random and independent

If that's not true, then the model would be wrong, I guess. As you say, a more "realistic" model might be hard, because it probably depends on homology between chromosome copies, which will probably vary between chromosomes, individuals, varieties, species, etc...

I suppose if the function was made general, this could be explicitly mentioned in the documentation and it would be up to the user to decide. Also, if the default ploidy = 2, then a substantial number of users don't even have to worry about this. 😄

[cfljam]

Im glad you are thinking about this. We have been trying QTLSeqR in some pooled data of hybrid backcross autotetraploids segregating for a major gene where the minor allele frequency of interest is 0.25. Results look sensible (similar to Popoolation CMH but sharper!) but Im interested how default expectations might not fit our system.

[WallyL]

Has anyone ever followed up on this? I would be very interested in learning what code modifications might be employed to facilitate better default modeling when running QTLSeqrR with a tetraploid species.