halasadi / ancient-structure

w/ Kushal
0 stars 0 forks source link

Prepare report for 11/22 Meeting #26

Closed halasadi closed 8 years ago

halasadi commented 8 years ago

(1) Show that Loschbour has two components and one of the two ("the green ancestry") is a better source population than the Loscbour ancestry (via log likelihood comparisons).

Loglikelihoods: https://github.com/halasadi/ancient-structure/blob/master/issues/issue21/loglik_values

Structure Plots:

The analysis was done on Moderns+Ancient pooled data but we plotted separately

Source: Yamnaya, LBK-EN, Loschbour

Moderns plot :

https://github.com/halasadi/ancient-structure/blob/master/issues/issue21/Admixture/modern_yam_lbk_losch/clus_3/struct_clus_3_.png?raw=true

Ancient plot:

https://github.com/halasadi/ancient-structure/blob/master/issues/issue21/Admixture/ancient_yam_lbk_losch/clus_3/struct_clus_3_.png?raw=true

Source: Yamnaya, LBK-EN, Green cluster

What is the green cluster here? Check first figure:

http://htmlpreview.github.io/?https://github.com/halasadi/ancient-structure/blob/master/exploratory/plot_exploratory.html

Moderns plot :

https://github.com/halasadi/ancient-structure/blob/master/issues/issue21/Admixture/modern_yam_lbk_green/clus_3/struct_clus_3_.png?raw=true

Ancient plot:

https://github.com/halasadi/ancient-structure/blob/master/issues/issue21/Admixture/ancient_yam_lbk_green/clus_3/struct_clus_3_.png?raw=true

Source: Yamnaya, LBK-EN, One unknown cluster

Moderns plot :

https://github.com/halasadi/ancient-structure/blob/master/issues/issue21/Admixture/modern_yam_lbk_unknown/clus_3/struct_clus_3_.png?raw=true

Ancient plot:

https://github.com/halasadi/ancient-structure/blob/master/issues/issue21/Admixture/ancient_yam_lbk_unknown/clus_3/struct_clus_3_.png?raw=true

(2). Show the scree plots and that: (i) there doesn't seem to be three dominant eigenvalues and (ii) the ancients have large variability

Check scree plot for both moderns and ancients

https://github.com/halasadi/ancient-structure/blob/master/exploratory/PCA/scree_pooled_and_moderns.png

Check scree plot for moderns only

https://github.com/halasadi/ancient-structure/blob/master/exploratory/PCA/scree_moderns_only.png

(3) Show the previous structure plots and remember that it was hard to visually distinguish between additional real source populations versus non-identifiable source populations. We showed that indeed, the modern data has extra clusters that doesn't get picked up by pooling the data.

For previous Structure plots, check:

http://htmlpreview.github.io/?https://github.com/halasadi/ancient-structure/blob/master/exploratory/plot_exploratory.html

Loglikelihood values of the comparison

https://github.com/halasadi/ancient-structure/blob/master/issues/issue23/loglik_values

(4) We pooled in additional data including Asian populations along with European populations to see how the new patterns show up.

We present the Structure plot from general admixture model fit on pooled ancient and modern data. Now we data 692 populations, with 620 modern and the rest being ancient populations.

K=7

Ancients: https://github.com/halasadi/ancient-structure/blob/master/issues/issue26/Admixture/ancient/clus_7/struct_clus_7_.png?raw=true

Moderns: https://github.com/halasadi/ancient-structure/blob/master/issues/issue26/Admixture/modern/clus_7/struct_clus_7_.png?raw=true

K=10

Ancients: https://github.com/halasadi/ancient-structure/blob/master/issues/issue26/Admixture/ancient/clus_10/struct_clus_10_.png?raw=true

Moderns: https://github.com/halasadi/ancient-structure/blob/master/issues/issue26/Admixture/modern/clus_10/struct_clus_10_.png?raw=true