Cross-coalescence rate query

[shaferab]

HI,

I'm having a difficult time interpreting my relative cross-coalescence rate in light of my MSMC2 plots and what I see online. It's raising a red flag, but wanted your take. Note, for ccr I used -I 0,4 and -P 0,1 combined; I've seen -I 0-4 as syntax but that is not accepted in my verssion. I've also followed the MSMC2 guideliine and the genome is decent. My Ne plots look similar to our PSMC runs on the genome, but we see a seperation happening >100 KYA (below), consequently the CCR is all over the map. This is a very dynamic species due to glaciations and multiple refugia but hoping for first thoughts / red flags from yourself or the community. Thanks ,

[stschiff]

Hi @shaferab, thanks for sharing. I think this CCR doesn't look too outrageous from what I've seen. I wouldn't over-interpret the small bump around 100kya, which never goes smaller than perhaps 0.85 and hovers there. I would rather take the time where the curve clearly passes from the upper half through to the lower half as a reasonable proxy for a separation time, so perhaps something like 30kya? Perhaps you can plot it with higher resolution within the last 150kya.

Also, for completeness, I recommend plotting the Ne-estimates from MSMC2, perhaps there is some evidence also from diverging Ne-estimates that support the population divergence seen in the CCR?

[shaferab]

Thanks. I put this in the wrong forum; all of this is with MSMC2. And thank you, that helps alot re: CCR interpretation and I will explore this further. 30KYA would align wiht glacial divergence. Interestingly, mtDNA from older work we did puts the N-S split at 225 kya.

[stschiff]

Ah OK... well, mtDNA are a single locus and can sometimes be non-representative. You should also consider phasing issues to potentially affect your results. Only simulations can help to assess that. It may not be necessary, but I have seen phasing switch errors leading to biased CCR curves.

I'll transfer this issue over to msmc2

[shaferab]

Thanks. One last follow up; could one interpretation of the NE plots AND full CCR being divergent lineages merging ~400 KYA, then seperating again 30 KYA?

[stschiff]

Perhaps, but your Ne plots are from PSMC, and the CCR from MSMC2, if I understand correctly. I would rather want to see the two from the same method to judge.

[shaferab]

Sorry, I wasnt clear - everything is msmc2 that is shown in the thread.

[stschiff]

Ah OK... Hmm, I don't think there is much support for a split around 400kya from the MSMC results. Yes, the Ne-curves look a bit divergent, but it's also clearly towards the right end of the resolution, and the CCR doesn't show that (the CCR going "up" as you move left on the right end of the plot is an artefact, there is no theory support for the CCR going up forward in time). I think the autosomal genome in this case supports a recent separation. Have you tried other methods, perhaps based on SFS, like momi2 or dadi to get independent on more model-based estimates of the divergence?

[shaferab]

Yes, dadi supports a very recent split, consistent with the CCR not mtDNA. I think, naively, I assumed a split would manifested in the Ne plots as well once they are diverged.

[shaferab]

Here's how it looks across pops zoomed in. Thanks again, can probably close now!

[stschiff]

Yeah, OK, looks pretty clear like a recent split to me. OK, I'll close, thanks for sharing!