Wednesday, November 14, 2012

STRUCTURE run on High/Low Altitude Ethiopians

The pdf can be downloaded here

Regarding the populations sampled, the paper notes the following:

The high altitude (HA) Amhara are agropastoralists living in a temperate Afro-alpine ecosystem in the Simien Mountains National Park at altitudes ranging from 3500-4100 meters (m). Altitudes above 2500m on the East African Plateau have been inhabited for at least 5 thousand years (ky) and altitudes around 2300-2400m for more than 70ky [24,25].”


DNA was extracted from blood samples provided by 192 Amhara individuals living at 3700 m in the Simien Mountains National Park or at 1200 m in the town of Zarima.”

For the Oromo:

The HA Oromo are pastoralists herding cattle, sheep and goats and living in a temperate Afro-alpine ecosystem in the Bale Mountains National park and reside on the Sanetti Plateau at 4000-4100m. The HA areas of the Bale Plateau have been inhabited by Oromo since the early 1500s according to historical records [22,23].”


79 individuals lived at 4000 m in the Bale Mountains National Park while 39 individuals lived at 1560 m in the town of Melkibuta.”

Melkibuta is probably a typo for Melkabuta, Bale, close to Goro, Bale which I have used as a proxy town in the map below for the location of the LA Oromo samples. 
Green= Low Altitude Amhara, Orange = High Altitude Amhara , Yellow = Low Altitude Oromo, Purple = High Altitude Oromo

Regarding the STRUCTURE run it says:

This position is further supported by the Bayesian clustering analysis performed using the program STRUCTURE [85]. In this analysis, 3 different sets of 57652 SNPs were used to infer the ancestral composition of each population assuming 7 ancestral groups. The STRUCTURE plots clearly show that Ethiopian populations share ancestral components with sub-Saharan African and Middle Eastern populations falling in the middle of the ancestry gradient between these two groups of populations (Figure S2.”

and Interestingly:

We also calculated the haplotype diversity and compared it to that observed in the worldwide populations. Interestingly, the Oromo (0.822) and Amhara (0.810) haplotype diversity values are as high as or higher than the highest values [80] observed in the HGDP, i.e. Bantu (0.818), Biaka Pygmies (0.815), Yoruba (0.815) and Mandenka (0.807); this is true regardless of altitude (0.798 for HA Amhara; 0.803 for LA Amhara, 0.813 for HA Oromo, and 0.813 for LA Oromo).”

There is also an FsT based Global neighbor joining tree in the PDF with a familiar outcome.

UPDATE: As far as the 7 clusters found in this global STRUCTURE run;
Cluster 1 (Blue) : Dominates in Sub-Saharan Africa, peaking in the hunter gatherers, AKA, Pygmy and Khoisan (Amhara ~28% , Oromo ~ 35%, Maasai ~ 56% )

Cluster 2 (Purple) : Dominates mostly in the Ethiopian and Maasai samples, but also found in North Africa, Near-east and West Africa in fairly significant amounts. (Amhara ~44% , Oromo ~ 44%, Maasai ~ 38% ) 

Cluster 3 (Green) : Dominates in West Asia / Europe, with a peak in the Sardinians. (Amhara ~28% , Oromo ~ 21%Maasai ~ 6% )

Cluster 4 (Orange) : Dominates in South Asia, peaking with the Gujarati samples.

Cluster 5 (Teal) : Dominates in East Asia.

Cluster 6 (Light Blue) : Dominates with Native Americans.

Cluster 7 (Brown) : Dominates with Oceanians.

Unfortunately the K=2 to 6 runs have not been reported, making it hard to gauge how this particular dataset would stack up relative to other global datasets.

UPDATE2: Comparing with ADMIXTURE.

Here, I compare the cluster breakdowns (or peaking populations) of the Global STRUCTURE run of this post, with the clusters formed in the ADMIXTURE global K=7 runs I have done in the past on two separate datasets, both datasets can be downloaded from here.

Dataset 1, Global, K= 7, Without Pagani 2012 East African Samples.

Cluster1: sardinian,basque,spaniards,italian,tuscans

Cluster2: dogon,yoruba,bambaran,hausa,igbo

Cluster3: irula,tn-dalit,ap-mala,ap-madiga,north-kannadi

Cluster4: san-nb,san,!kung,pygmy,mbutipygmy

Cluster5: papuan,melanesian,tongan,samoan,paniya

Cluster6: colombian,surui,karitiana,pima,totonac

Cluster7: she,han,chinese-americans,singapore-chinese,chinese

Dataset 2, Global, K= 7 , With Pagani 2012 East African Samples.

Cluster1: papuan,irula,tn-dalit,ap-mala,ap-madiga

Cluster2: sardinian,basque,spaniards,italian,tuscans

Cluster3: san-nb,san,!kung,pygmy,mbutipygmy

Cluster4: colombian,karitiana,surui,pima,totonac

Cluster5: she,chinese,han,chinese-americans,singapore-chinese

Cluster6: yoruba,dogon,brong,igbo,bambaran

Cluster7: ARI-B,ARI-C,Gumuz,Somali,EtS-P

The clusters highlighted in yellow are clusters also found by the STRUCTURE run of this post (at least roughly), the main differences are in the African clusters, while ADMIXTURE split the African clusters between a West African, Hunter Gatherer and East African (only in the case of the Pagani inclusive samples), the STRUCTURE run did not find 3 but rather only 2 African components, instead, as a compensation it split the Oceanians from the South Asians. There can only be three explanations to these differences in results:

  1. The SNPs used are from different regions of the genome
  2. The way STRUCTURE splits components is different from ADMIXTURE
  3. The difference in sampling of the global datasets (of which those of the ADMIXTURE runs were more complete)

Or all 3 could be true with varying degrees of impact. The only way to verify is by running ADMIXTURE with a global dataset similar to the one in this post.

No comments:

Post a Comment