Stdpopsim: More complex admixture graph model

[jeromekelleher]

Want an example of a more comple admixture graph model. Can you provide a link to an appropriate reference please @FerRacimo?

May depend on #136 though.

[FerRacimo]

I was just talking to Aaron about this and there's two definitions of
complexity here.

A complex topology of an admixture graph would be something like Lipson and
Reich (2017) Figure 1 or Sikora et al. (2019) Extended Figure 3 (links
below). However, for these cases, we would need to come up with our own
population sizes and times for each branch of the graph (such that they
result in the same amount of drift).
https://academic.oup.com/mbe/article/34/4/889/2838774/
https://www.nature.com/articles/s41586-019-1279-z

A complex multi-population model with fully specified times and effective
population sizes would be something like Excoffier et al. 2013 Figure 4 or
Kamm et al. 2018, Figure 3 (links below). Here, the topology would not be
complicated (most fully specified model inference methods like momi2,
fastsimcoal2 or moments don't go beyond more than ~10 populations at a
time) but we would at least have the full list of parameters, so we
wouldn't need to invent any:
https://journals.plos.org/plosgenetics/article?id=10.1371/journal.pgen.1003905
https://www.tandfonline.com/doi/full/10.1080/01621459.2019.1635482

I think we can do both but the second option would be more straightforward
than the first. I'll start with the second thing once I'm back from Aussois.

On Tue, Oct 8, 2019 at 6:09 PM Jerome Kelleher [e-mail]
wrote:

Want an example of a more comple admixture graph model. Can you provide a
link to an appropriate reference please @FerRacimo
https://github.com/FerRacimo?

May depend on #136 https://github.com/popgensims/stdpopsim/issues/136
though.

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
https://github.com/popgensims/stdpopsim/issues/140?email_source=notifications&email_token=ABLXSAUPEMWLJX2PX4BSIYLQNSWCZA5CNFSM4I6T3BAKYY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4HQMRYDQ,
or mute the thread
https://github.com/notifications/unsubscribe-auth/ABLXSASKL7S72PGL2VSWUGLQNSWCZANCNFSM4I6T3BAA
.

--
Fernando Racimo, PhD
Assistant Professor
Centre for GeoGenetics
University of Copenhagen
Øster Voldgade 5-7, DK-1350 Copenhagen, Denmark

[FerRacimo]

So I just finished coding the Kamm et al. model in my own branch but have an issue. The model includes some ancient samples with fixed dates in years. At the moment, from what I can see, the get_samples function can only take time=0 samples. One could modify this but, ideally, it would be even better if a model creator could fix the times from the model itself (in a similar fashion to how one can specify allow_samples=False for ghost populations). I'm not familiar enough with the code to this myself though. Could someone take care of this?

[jeromekelleher]

I had a look at the PR @FerRacimo, and it looks great. I think we can treat the sampling thing as a separate issue though: we can still take samples from the present day, and figure out the right way to take samples from the ancient populations afterwards. Does that sound OK?

[FerRacimo]

Sure, I'll submit the pull request then. Just keep in mind that, if a user
is free to run the Kamm et al. model without ancient sampling, it will have
weird features: e.g.: a Neanderthal artificially sampled in the present
after 120,000 years of having Ne = 100 will be tons more inbred than the
actual Neanderthal genome that was used to fit the model (which is already
quite inbred). The model also depends on having obtained samples from
ancestral populations before particular admixture / divergence events
happened, so this may also lead to artifacts.

On Tue, Oct 15, 2019 at 10:48 AM Jerome Kelleher [e-mail]
wrote:

I had a look at the PR @FerRacimo https://github.com/FerRacimo, and it
looks great. I think we can treat the sampling thing as a separate issue
though: we can still take samples from the present day, and figure out the
right way to take samples from the ancient populations afterwards. Does
that sound OK?

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
https://github.com/popgensims/stdpopsim/issues/140?email_source=notifications&email_token=ABLXSAV6R3MJFKUQBPSP3PTQOV7ULA5CNFSM4I6T3BAKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEBH6S6I#issuecomment-542108025,
or unsubscribe
https://github.com/notifications/unsubscribe-auth/ABLXSAXAL3DPQFOWPBKAFITQOV7ULANCNFSM4I6T3BAA
.

--
Fernando Racimo, PhD
Assistant Professor
Centre for GeoGenetics
University of Copenhagen
Øster Voldgade 5-7, DK-1350 Copenhagen, Denmark

[jeromekelleher]

Closed in #146