apragsdale/african-structure-paper

A weakly structured stem for human origins in Africa

The fit models presented in this paper are available in Demes format as YAML files, found in the demes_models directory in this repository.

This paper is posted as a preprint at bioRxiv: https://doi.org/10.1101/2022.03.23.485528.

Contents

This repository contains the manuscript and supplementary material for "A weakly structured stem for human origins in Africa." To build the document, simply type make in this directory. If you have only updated the supplement, you may need to type make -B instead to force a rebuild.

GitHub is set up to automatically build the manuscript from the most recent commit, and PDFs of the paper and supplement can be found here.

Getting started

You may need to install pdflatex to compile the document.

On linux:

sudo apt-get install texlive-latex-base texlive-fonts-recommended texlive-fonts-extra texlive-latex-extra

On mac (untested):

brew install basictex

Contributing

You can make small edits in the browser by navigating to the file and clicking the "edit" button (which looks like a pencil in the top-right corner of the window). Any edits you make can then be committed directly to the repository by completing the "Commit changes" section at the bottom of the page. Just include a short phrase saying what you edited and click the green "Commit changes" button.

To make larger edits and to compile the document, you should first clone the repository to your own computer, make changes there, commit the changes locally, and then push those changes back to the repository.

Note: It's best to make small commits and push often to ensure conflicts don't arise. Also, it's helpful to coordinate to make sure only one person is tackling a given section at a time to avoid merge conflicts.

To clone the repo

cd Documents  # or wherever you want this project to live locally
git clone [email protected]:apragsdale/african-structure-paper.git
cd african-structure-paper
make

To push your recent edits to the repo

After editing some section, do the following

git add [files-you-edited]
git commit -m "very short message describing the edits"
git push origin main

Making sure your copy is up to date

If you've already cloned the repository, you want to make sure you are editing the most up-to-date version of the document. Before doing any editing, pull changes from the repository:

git pull origin main

What to do in case of conflicts

The best way to avoid conflicts is to always make sure you are working with the most up to date version of the document, by pulling from the repo (see section above). But it can sometimes happen that someone else has pushed changes to the repo in the time between pulling and pushing changes yourself. In that case, pushing the repo might generate a message that says something like

 ! [rejected]       main -> main (fetch first)
 error: failed to push some refs to `github.com:apragsdale/african-structure-paper.git`

as well as some extra lines that have a hint.

To be able to push the changes in this case, you need to pull and merge the more recent changes from the repo:

git pull origin main

With any luck, everything merges properly, and you can try pushing againg

git push origin main

african-structure-paper's Issues

Discussion of SMC curves

The second paragraph in "Reconciling multiple lines of genetic evidence" currently reads:

Similarly, we generated estimates of effective population size as inferred from coalescence rates with Relate. Changes in Ne over time from prior studies strongly support a bottleneck in Out of Africa populations, but African populations also dip in effective size around 200kya, although this decrease is more modest than the OOA event [Mallick 2016]. One possible explanation is that ancestral population structure during the Middle Stone Age inflated Ne and subsequent gene flow among ancestral Africans resulted in a lower Ne after 100-200kya [Chikhi?]. To our knowledge, explicit comparisons of this hypothesis with SMC-style approaches have not been published. We simulated genomic data under 4 models (Figure SX). All models, including the Single Origin model, generate multi-modal curves indicating that the ancestral increase in Ne between 100kya-1Mya is a general feature of the coalescent rate in humans. Simulated Relate Ne curves under the CM and MM are distinguished by having 2 and 3 modes, respectively. The Continuous Migration model had the better qualitative fit with the data, but the ancestral mode was shifted further back in time relative to the data. All simulated models underestimated the level of population growth during the Holocene, indicating additional free parameters are necessary to refine these growth rates.

There has been some discussion here, pasted from the google doc:

Simon Gravel
4:03 PM Aug 18
This can be much improved. It would be good to speak to inverse coalescence rates as well, and s*

Simon Gravel
4:55 PM Dec 21
On S*, should we just say that the statistic was not used directly in quantitive fits?

Simon Gravel
4:56 PM Dec 21
Also, it would be nice to have one interpretative comment on the relate curve, otherwise it is not clear why we point this out.

Brenna Henn
2:17 PM Dec 30
OK, I've written a couple of paragraphs. I don't think know if this is what you had in mind - but is my reading of the Relate plots. We could move some of this to Supp.

Brenna Henn
2:23 PM Dec 30
Personally, I think the qualitative fit with Relate curves is pretty poor.

Brenna Henn
2:23 PM Dec 30
If anything, the single origin model best resembles the data

Brenna Henn
4:03 PM Dec 30
I generally think CCR are not helpful for reconstructing demographic history. You may also be interested in seeing our Standard PopSim OOA run thru Relate https://www.dropbox.com/s/2rjrzjax7b3caew/StdPopSim_OOA_Relatecurves.pdf?dl=0

Aaron Ragsdale
3:58 PM Dec 31
I agree that coalescence rate curves may not be helpful for reconstructing demographic history, especially at points farther back in time. But they are still something that can be estimated from data and compared to the model - it's the interpretation of those curves as population size history or drawing inferences of population splits from them that is a problem. The IIRC and CCR curves are still "real" and can be compared.