Drosophila recomb rates should be divided by two!

[petrelharp]

In Drosophila, males don't do recombination; and AFAICT the rates we have are untransformed from Comeron et al, which are reporting crossovers per female meiosis.

However, our simulations have recombination happening in all individuals.

A better alternative would be to implement "females-only" recombination, which would not be that hard (it would require an additional species-level property, though). But either way the first question is: is this right, that we're getting twice as much recombination as we should?

[andrewkern]

i think what we are doing is correct because Comeron estimated the number of recombinations in a cross. If males were also to recombine Comeron's estimates would be ~2x higher.

[dschride]

As far as I can tell from the phrasing in the Comeron et al. paper, they are reporting the rates per female meiosis.

[petrelharp]

I think we're all in agreement - reported numbers are number of recombinations in a cross (which = number of recombs per female meiosis). So, are we in agreement that this means wer'e simulating twice as much recombination as we should?

[RyanGutenkunst]

My group made the same mistake years ago using the Comeron map (which a reviewer caught!). The do report rates per female meiosis, so if the map is applied to all individuals, it is twice as much recombination as there should be.

[andrewkern]

okay! this is a great catch @petrelharp

[petrelharp]

Okay - here's my proposal for how to fix this:

1. divide the rates by two and add a prominent message explaining this to the docs for them
2. put on the TODO list to do stuff with sexes appropriately

Alternative route:

1. leave rates as-is (with prominent message)
2. hack something in temporary that checks if it's Drosophila and if so, divides by two
3. longer-term, add a property that says "sexes recombination occurs in" and uses that property appropriately

Advantages to dividing by two:

• rates are correct for msprime and non-sexual SLiM models
• definition of rates agree with other species (they are sex-averaged)

Advantages for leaving as-is:

• rates are correct for the arguably-more-correct to-be-implemented-recomb-only-in-females SLiM model

I think it's pretty clear then that we just want to divide all the rates by two.