Clusters Galore analysis of East African participants

See here.

fastIBD analysis of Italy/Balkans/Anatolia

I have included the new Turkish data from Hodoğlugil & Mahley (2012) in this analysis. Additionally, there are now 5 participants in the Serb_D and Turkish_Cypriot_D sub-populations, as well as a Bosnian Muslim. There are now project participants from many Balkan countries, although Albania, the fYROM, and Croatia remain as "black holes" in the map.

Still, I am hopeful that there will be more project participants from currently under-represented populations. I have already started processing the same dataset with ChromoPainter (which takes much longer), and hopefully that analysis will be posted at the end of this week or the beginning of the next one.

First, the heatmap of inter-population IBD:

Remember that the tree groups similar populations together, and for each row in the matrix, the red end of the spectrum indicates lots of IBD sharing, and the blue end low IBD sharing. Additionally, I have now calculated the median IBD sharing, which is more resistant in the presence of potential relatives in the data.

The results appear fairly reasonable, with the Balkan, Anatolian, and Italian populations of the title forming separate branches, and the mainland Greek sample joining with Central/South Italians and Sicilians.

The Clusters Galore can be seen below; 28 clusters were inferred with 21 dimensions:

Results for Project participants can be found in the spreadsheet, and include the probabilities that each ID is assigned to each of the 28 clusters, as well as the Z-scores comparing each individual against all populations with 5+ individuals. The Z-score should be read as follows: for each row, high values indicate a high degree of IBD sharing, while low values indicate a low degree of IBD sharing.

Of course, I encourage Project participants to leave a message in the Information about Project samples thread.

fastIBD analysis of Afroasiatic groups (Jews, Arabs, Assyrians, Berbers, Somalis, Amharas, etc.)

Please refer to the previous analysis on the Balkans/West Asia for more information about the interpretation of this type of analysis.

I am very pleased with the way this analysis of Afroasiatic groups has turned out, revealing an exceptional degree of resolution. I invite individuals from the Near East and Africa who are eligible, to submit their data, so that they can be included in future runs of this kind.

Clusters Galore


45 clusters were inferred with 29 dimensions.

I can't comment on all 45 clusters, so I'll just limit myself to the ones that are significantly represented among Project participants: 1. Ashkenazi, 4. Assyrian/Mandaean, 6. Somali, 7. Moroccan, 8. Algerian/Tunisian, 9. Sephardic, 10. Morocco Jews, 11. Iran/Iraq Jews, 12. Non-Jewish Ethiopians, 13. Saudi, 14. Arab #1, 15. Arab #2, 16. Egyptian

Inter-Population IBD

Results for Project Participants


The results can be found in the spreadsheet.

I have also added the full IBD sharing matrix which lists how many Morgans of sequence are estimated to be IBD with probability greater than 10^-6 between all pairs of individuals.

You can google any non-Project sample IDs to get some more information about their origin. For example, GSM536710 is an Iraqi Jew who shares about half his genome with GSM536714, also an Iraqi Jew. These two samples are almost certainly first-degree relatives. Or, GSM537032, a Samaritan shares 740-1,480cM with the other 2 Samaritans, an exceptional amount in this small and probably highly inbred population.

You can manipulate this matrix in R. After you download it and unzip it, you can load it into R as follows:

X<-read.table('afroasiatic_ibd_sharing.txt',row.names=1,header=T)

Then, you can, for example, sort the IBD sharing for a particular individual, as follows:

sort(X['DOD026',])

fastIBD analysis of Central/Eastern Europe

Please refer to the previous analysis on the Balkans/West Asia for more information about the interpretation of this type of analysis.

Clusters Galore


The Clusters Galore can be found in the spreadsheet. After inspection of the 23 clusters inferred with 21 dimensions, they could be described as:

1. Mordvin
2. East Slavic
3. Polish-Ukrainian
4. East Balkan
5. Vologda Russians
6. Lithuanian
7. Central European (combining many groups with small sample sizes)
8. A couple of related (?) individuals
9. Anatolian
10. Greek
11. Chuvash
12. Ossetian
13. A couple of related individuals
14. A couple of related individuals
15. Balkar
16. A couple of related individuals
17. Chechen
18. Kumyk
19. A couple of related individuals
20. Adygei
21. Lezgin #1 (main)
22. Lezgin #2
23. Lezgin #3

If you belong to a population with few other participants, you might end up latching onto a cluster dominated by a bigger group. This does not mean that your population is not distinctive, only that there are not enough samples to reveal its distinctiveness if it exists.

Inter-Population IBD

Results for Dodecad Participants

Results can be found in the spreadsheet.

If you have joined the Project, please consider leaving a comment in the Information about Project samples thread. That will help others make better sense of their results, e.g., if you find that you belong in the same cluster with some other individual, you might want to know something about their origins.

UPDATE: I have added the IBD sharing matrix.See here on how to use it.

fastIBD analysis of South Asia

Please refer to the previous analysis on the Balkans/West Asia for more information about the interpretation of this type of analysis.

Clusters Galore


The Clusters Galore analysis can be found in the spreadsheet. 59 clusters were inferred with 47 MDS dimensions. The very fine-scale structure (I only considered the first 50 dimensions, but many more seemed significant than in any previous experiment) is probably the result of the size of the South Asian population, as well as the practice of endogamy associated with the caste system. High intra-population IBD sharing is also evident in the following (notice how well-defined the diagonal is):

Inter-Population IBD

Results for Dodecad participants

They can be found in the spreadsheet. Many Project participants belong to a population with 1 or 2 individuals, so cluster #1 seems to be a generalized catch-all for many such individuals. Individuals from he two sub-populations that I've identified recently Iyer_D, and Jatt_D all belong to the same cluster. The Iyer_D cluster (#4) also seems to include the Iyengar project participants as might be expected.

It is also interesting how all Dodecad participants fall in just 7 of the 59 clusters. This goes to show how truly diverse people from the Indian subcontinent are. I fully expect that with more participation further structure will be revealed, since it seems that due to endogamy it only takes a few participants from each ethnic group for a specific cluster pertaining to that group to be identified. So, I invite people from South Asia to join the Project during this submission opportunity.

fastIBD analysis of Iberia, France, Italy, Balkans, Anatolia and European Jews

On the heels of the previous analysis of Balkans/West Asia, a new experiment on a different set of populations. Please refer to the earlier post for some thoughts/explanations about this type of analysis, I'll stick to "just the data" for this post.

Clusters Galore

24 clusters inferred with 17 MDS dimensions.

The Galore analysis provides increased resolution within Iberia (#6-9, 11), Italy, and the Ashkenazi Jewish group (#14-16).

The Iberian results are particularly interesting, showing the power of this approach compared to the one with unlinked data. There appear to be:

• a Spanish Basque (#6), 
• French Basque (#11) cluster, as well as 
• a Portuguese/Galician/Castilla Y Leon (#9) cluster, and 
• a complementary Castilla La Manch/Cantabria/Andalucia/Murcia (#7) cluster, and 
• a smaller Aragon/Cataluna cluster (#8). 

There is overlap between these clusters, but the geographical contrasts are quite evident. I did not go through the results of Spanish Project participants (all the Portuguese fall in the Galician cluster, and our Basque member in the Basque cluster as expeccted), so it would be interesting to hear whether they fall in the cluster(s) which exist in their regions of origin.

Inter-Population IBD

Results for Project Participants


The results can be found in the spreadsheet.

fastIBD analysis of Balkans/West Asia

Now that I've discovered a way to boost Clusters Galore analysis even further by using fastIBD, I will start experimenting with different regional populations. This analysis took about 5 hours to complete, so it appears to be quite practical.

For my first experiment, I carry out an analysis of various populations from the Balkans and West Asia.

Clusters Galore

27 different clusters were inferred with 17 MDS dimensions. Some interesting findings:

• For the first time there emerge a couple of clusters that appear to be quite specific to Armenians (#2 and #3). 
• Similarly, Assyrians are broken to a few clusters that appear fairly specific to them  (#9-11)
• Georgians are split into three clusters, one of which (#14) is linked with the neighboring Abkhasians, who in turn have their own exclusive cluster (#25)
• The cluster modal in Greeks (#6) includes 14 of 19 Greek participants, and a few Greeks are also in the Balkan cluster (#8) and an Iranian-Turkish cluster (#4)
• The Behar Cypriot sample also splits into two, and the few Turkish Cypriot participants link to one of them (#13)
• The Ossetian project participant links to one of the three North_Ossetian clusters
• The major Balkan cluster (#8) still defies resolution. I am certain, however, that structure in this cluster will be uncovered with more participation. MCLUST adapts the cluster size and shape, and a "big", inclusive cluster spanning the Balkans appears more parsimonious than smaller clusters centered on the different groups. With larger participation, I anticipate that regional structure will be uncovered in the Balkans as well.

I cannot stress the importance of participation strongly enough. When groups have more participants, it is possible to both:

1. Discover group-specific clusters, by identifying what is common between members of groups
2. Discover within-group clusters, by identifying what is different between members of groups

For example, the great participation of Armenians in the Project has now allowed me to discover structure within the Armenian population. It appears, that cluster #2 corresponds to a more "western" Armenian group, and #3 to a more "eastern" one, with some overlap between the two.

Inter-population IBD


You can also see a visual representation of inter-population IBD:

I have only included populations with 5+ participants in this representation. Reddish shades express high IBD sharing; bluish ones low one. The heatmap has been scaled by row.

As you might expect, values across the diagonal are "reddish", since individuals within populations tend to have high IBD sharing with each other.

A few features "pop out" of the screen. Going from top to bottom:

• Intra-Iranic sharing
• Intra-Armenian sharing
• Intra-Balkan sharing
• Georgian-Abkhaz sharing

You can probably get more out of the figure, but these appear to be the most salient features.

Results for Project Participants


The results can be found in the spreadsheet, and include:

• Probabilities of assignment in each of the 27 clusters of the Clusters Galore analysis
• Z-scores of IBD between each individual and each of the 20 populations with 5+ participants. Higher values mean more IBD sharing. Note that Z-scores have been calculated for each row, hence each participant must scan his own row to find populations with an excess (+) or deficiency (-) of IBD sharing, and people should not compare across different rows.

Last but not least, I want to remind new project participants to leave a message in the Information about Project samples thread. Your comment will not appear immediately, since comment moderation is on, and also note that there are multiple pages of comments. 


If you haven't joined the Project yet, I encourage you to do so if you are eligible.

Clusters Galore (fastIBD edition) for some northern European participants

You can find some new Clusters Galore results here (scroll down for spreadsheet link). The new methodology described in that post has made it possible to infer even finer-level population structure than "classical" Clusters Galore.

Clusters Galore: African edition

I gathered about 900 individuals from the Dodecad Project and the literature and run Clusters Galore on them. Dodecad Project members were included if any of the following conditions held:

1. They were in the North_African_D, North_African_Jews_D, or East_African_D population
2. They had at least 10% "Northwest African" in the standard K=10 analysis
3. They had at least 10% "West African" in the standard K=10 analysis
4. They had at least 10% "East African" in the stanard K=10 analysis

Integrating among multiple datasets meant that the analysis was done on only 1,871 SNPs, which were however sufficient to infer 28 clusters with 9 PCA dimensions retained.

PCA plots

Below are PCA plots of the first nine dimensions; each dimension is paired with the 1st one to create 8 different scatter plots:

Galore/PCA results

All the results can be found in the spreadsheet, which contains:

• Population Galore results: how many individuals from each population are assigned to each cluster
• Individual Galore results: what is the probability that each ID belongs in each cluster (in %)
• Population PCA results: mean positions of populations in the first 20 principal components
• Individual PCA results: position of individuals in the first 20 principal components.

Some observations

It does appear that within Africa two processes explain most of genetic variation:

1. Variable affiliation with West Eurasians, with North Africans being most West Eurasian-like, followed by some East Africans such as Ethiopians, and then Maasai
2. Contrast between farmers and hunter groups such as the San and Pygmies

With respect to Project participants:

1. All members of East_African_D align with Ethiopians and Ethiopian Jews (cluster #2)
2. All members of North_African_Jews_D align with a major cluster of the Morocco_Jews of Behar et al. (2010) (cluster #3)
3. Members of North_African_D are split over three clusters: most are in the Algeria/Libya/North Morocco cluster #4; one is in the Egyptian cluster #5; three are in the Mozabite/Saharan/"Berber" cluster #7
4. The Other_D participants include many African Americans, as well as some Arabs with some African admixture, etc. Clusters #8 and #10 seems to include many African Americans.

Hopefully this will be useful to project partipants with an African background.

Structure in West Asian Indo-European groups (part 2)

I will occasionally revisit old posts such as Structure in West Asian Indo-European groups to take advantage of new population samples from project submitters. This time around, I included our first Kurdish and Azeri participants, and limited the analysis to populations for which I had large numbers of markers (the final total is a ~132k pruned set of markers).

I also included Greeks, Caucasian populations (Georgians and Lezgins), and Levantine Arabs (Syrians-Lebanese) who frame this region from the West, North, and South respectively, as well as Assyrians who are interspersed in West Asia as an ethno-religious minority.

Here are dimensions 1 and 2 of the multidimensional scaling plot:

The two Caucasian groups (South and Northeast Caucasian Georgians and Lezgins respectively) form their own clusters. So do the Iranians and the Syro-Lebanese.

(As always population labels are placed on population averages, and _D denoted Dodecad Project populations)

Curiously, many linguists assert a close relationship of Greek and Armenian within the Indo-European language family. Turks speak an Altaic language due to migration of a numerically small population element, but their pre-Turkish genetic ancestors were Anatolian speakers, Greeks, Armenians, and Iranians, i.e., primarily Indo-Europeans.

Dimensions 1 and 3:

Dimension 3 contrasts Greeks from West Asian groups. Notice also the presence of 3 Armenians at the bottom of the plot, these are outliers of the Behar et al. Armenian sample.

Notice that the Azeri_D sample clustered with Iranians in dimension 2 and with Turks in dimension 3. This is not very surprising, as Azeris speak a Turkic language, but also have clear Iranian antescendants. The Kurd_D sample, on the other hand, clusters consistently with Iranians.

The variability of the Greek_D population sample along dimension 3 is also quite interesting. This could reflect variable levels of influence of extra-Greek European/Anatolian population elements on the basic Greek stock. Greeks who possess 23andMe or FTDNA Illumina population data are strongly encouraged to join the Project to help us better determine regional variation within the ethnic Greek population.

The Clusters Galore analysis results are as follows (9 clusters/3 MDS dimensions retained) :

In brief, the modal populations in each cluster are:

1. Greeks
2. Iranians
3. Turks
4. Syrians and Lebanese
5. Armenians and Georgians
6. Armenians and Assyrians
7. Syrians and Iranians
8. Lezgins
9. Georgians

I will be happy to provide to all Dodecad Project members from the _D populations with their individual results. If you send me e-mail at dodecad@gmail.com, I will send you a line with your probabilities of assignment in each of the 9 clusters, as well as the 3 co-ordinates in the first 3 MDS dimensions plotted above.

I strongly encourage individuals from West Asia, the Balkans, and Italy to contact me for inclusion in the Project (send e-mail first, not data). While submission to the Project is currently closed, I usually accept data from these regions.

Clusters Galore results, K=73 for Dodecad Project members (up to DOD581)

The results can be found in the spreadsheet. There are 73 clusters with 13 MDS dimensions retained. The spreadsheet contains 558 rows for unrelated project participants, each of which contains the probabilities that each individual belongs to one of the 73 clusters. This is followed by 36 rows for the reference populations, showing how many individuals from each population are assigned to each cluster.

In order to interpret your results, first search for your DOD number, and see which columns you have non-zero probabilities in. For the vast majority of individuals you will be uniquely assigned (100%) in one of the 73 clusters. Then, you can visit the ancestry thread to see who else is assigned to the same cluster as yourself, and also look in the reference populations to see how they are represented in the different clusters.

The following 67 IDs were characterized as outliers:

DOD002 DOD004 DOD006 DOD020 DOD029 DOD030 DOD034 DOD036 DOD060 DOD063 DOD072 DOD075 DOD081 DOD107 DOD126 DOD128 DOD132 DOD133 DOD156 DOD157 DOD168 DOD169 DOD175 DOD183 DOD201 DOD224 DOD234 DOD245 DOD252 DOD303 DOD309 DOD316 DOD326 DOD328 DOD339 DOD348 DOD349 DOD359 DOD363 DOD380 DOD382 DOD385 DOD388 DOD392 DOD393 DOD422 DOD425 DOD430 DOD435 DOD437 DOD489 DOD492 DOD495 DOD500 DOD502 DOD511 DOD521 DOD523 DOD531 DOD533 DOD536 DOD548 DOD571 DOD572 DOD573 DOD574 DOD577

As previously explained, outliers may either be mixed individuals or individuals from particular populations not well represented in the Project. In both cases they appear to be more "distant" from other individuals and from their respective clusters.

Getting back to the 74 inferred clusters:

• Cluster #2 is by far the largest, consisting of mainly of "British Isles"/American White types of people; this grew substantially because of the recent open submission call when many people of this type of ancestry joined
• Cluster #3 is essentially Ashkenazi Jewish, another big group in the Project
• Cluster #5 is not represented in the reference populations except for a single Utah White. This is largely German.
• Cluster #6 is mostly (but not exclusively) French.
• Cluster #9 is largely Finnish and also includes some East Slavs.
• Cluster #12 is essentially South Italian/Sicilian/Greek
• Cluster #14 is mostly Assyrian/Armenian
• Cluster #16 is mostly Balkan
• Cluster #21 is mostly Scandinavian
• Cluster #27 is mainly Balto-Slavic
• Cluster #29 is essentially Iberian

I covered most of the largest clusters, but there are also plenty of smaller ones. So, make sure you contribute/read the ancestry thread to get a feel for the kind of people that share your cluster. Many of mixed-race participants (e.g., African Americans) are split into multiple clusters; I recently observed that this is the case for highly variable populations with inter-continental admixture.

Don't forget also, that sharing a cluster does not imply a very strong genetic similarity, as clusters may be either very tight or very loose. This analysis is better at identifying differences than at confirming strong similarities.

Readers of the blog will be aware that many of these clusters can be subdivided further if a regional analysis is carried out (e.g., the Assyrian-Armenian one), while others have proven difficult to split meaningully (e.g., the Iberian one into Spanish vs. Portuguese).

This time around, I included all the Genomes Unzipped people as well as Lily and Greg Mendel (LIL001, GRM001).

I will be exploring these clusters further, and any further regional structure that I may discover will be posted in this blog. So, do subscribe to the feed as there may be additional results for your sample ID.

Clusters Galore results, K=63 for Dodecad Project members (up to DOD449)

The results can be found in the spreadsheet. There are 63 clusters with 11 MDS dimensions retained. The spreadsheet contains 409 rows for unrelated project participants, each of which contains the probabilities that each individual belongs to one of the 63 clusters. This is followed by 36 rows for the reference populations, showing how many individuals from each population are assigned to each cluster.

In order to interpret your results, first search for your DOD number, and see which columns you have non-zero probabilities in. For the vast majority of individuals you will be uniquely assigned (100%) in one of the 63 clusters. Then, you can visit the ancestry thread to see who else is assigned to the same cluster as yourself, and also look in the reference populations to see how they are represented in the different clusters.

The following IDs were outliers:

DOD004 DOD006 DOD010 DOD020 DOD024 DOD029 DOD030 DOD032 DOD036 DOD047 DOD050 DOD051 DOD053 DOD060 DOD063 DOD072 DOD075 DOD107 DOD126 DOD128 DOD132 DOD133 DOD156 DOD157 DOD168 DOD169 DOD175 DOD216 DOD223 DOD235 DOD239 DOD240 DOD245 DOD252 DOD294 DOD303 DOD316 DOD326 DOD339 DOD348 DOD359 DOD363 DOD380 DOD382 DOD384 DOD385 DOD387 DOD388 DOD389 DOD425 DOD430 DOD435

As previously explained, outliers may either be mixed individuals or individuals from particular populations not well represented in the Project. In both cases they appear to be more "distant" from other individuals and from their respective clusters.

A few observations:

• The single largest cluster is #3 which is mostly "British Isles"
• Cluster #26 encompasses most Greek/South Italian/Sicilian individuals; not how this is not represented in the reference populations, which lack such individuals
• Cluster #23, also absent in the reference populations encompasses mainly Finns and some Russians
• Cluster #25 is also quite large, consisting of 42 Project members and only 2 reference White Utahns. This consists largely of North/Central Europeans from continental Europe.
• Cluster #4 includes mainly Iberians
• Cluster #11 mainly Ashkenazi Jews
• Cluster #15 mainly Turks
• Cluster #16 mainly people from the Balkans
• Cluster #27 mainly North-Central Italians not in #26 (the Greco-Italian cluster)

As always, I encourage those who haven't posted in the ancestry thread yet to do so, to help themselves and others make better sense of their results.

I plan to explore fine-scale structure of Dodecad Project members further, especially of those who belong to large, undifferentiated clusters that may harbor latent informative structure.

Clusters Galore with Dodecad populations

Here is a spreadsheet of Clusters Galore analysis with Dodecad populations: 692 reference individuals + 261 Dodecad Project participants from 24 different populations with at least 5 members each:

Assyrian, Scandinavian, Greek, Finnish, S_Italian_Sicilian, Ashkenazi, German, Indian, Portuguese, Armenian, Russian, Spanish, British, Irish, Turkish, N_Italian, Balkans, Iranian, North_African, East_African, French, Chinese, Japanese, Polish

As a reminder to new readers, the Clusters Galore technique consists of applying multidimensional scaling on genomic data to convert ~152,000 SNPs into a number of continuous dimensions capturing most of the variation, followed by employment of MCLUST to cluster individuals along these dimensions.

In total, 65 clusters were obtained when 10 MDS dimensions were retained.

Some observations:

• Most Greeks and all South Italians/Sicilians continue to fall in the same cluster #4. The fact that the latter population, despite being one of the largest (20 individuals) continues to remain unsplit and distinctive testifies to the fact that it is probably homogeneous and lacks substantial regional inbreeding within it.
• Cluster #2 includes most Germanic individuals and also the Irish
• Cluster #5 is made mostly of Central/North Italians
• Non-Greek Balkan participants fall mostly in cluster #6, which also includes the non-Gypsy admixed reference Romanians
• Project and reference Iberians (Spaniards and Portuguese) continue to be undifferentiated and distinctive, falling in cluster #14; my comments on South Italians/Sicilians also probably apply to them as well.
• There is a trace of structure in the Ashkenazi population, which is split into two clusters. This probably underscores the benefits of large samples in the inference of structure, as 25 Ashkenazi Jews have submitted their results to the Project.
• Project Russians have split affiliations between a circum-Baltic cluster #3 and the Finnish cluster #7.
• North Africans form two new clusters that do no overlap with either reference Mozabites or Egyptians. There is great variety in North Africa, and the 8 people who have submitted their samples are a good start to learning about this region of the world.
• The Chinese are split into two, one part aligning with the "southern" Miaozu and one part aligning with the "northern" Japanese.

Please do not ask me which cluster you fall in, as there will be a separate analysis of Project participants identified by their DOD number but without ethnic identifiers, in compliance with the Project's privacy policy.

Clusters Galore results, K=64 for Dodecad Project members (up to DOD307)

The results are found in the spreadsheet. 64 clusters were inferred with 12 MDS dimensions retained. Check the previous K=56 analysis and links therein to familiarize yourself with this ancestry inference technique.

The results table consists of 271 rows for non-related Project members with 23andMe data, each of which contains the probabilities that they belong to one of the 64 clusters. This is followed by a table in which the number of individuals from each reference population assigned to each cluster is shown.

The increase in the number of individuals, compared to the previous analysis (up to to DOD236) has resulted in the inference of a greater number of clusters. Once again, I recommend that people leave a comment in the ancestry thread because that is the only way in which you will find out where people in your cluster are from.

I did not fully inspect all the clusters to see what each of them corresponds to. Some observations:

• #8 seems to be the mainly Greek/South Italian cluster previously detected
• #3 seems to be a Finno-Ugrian cluster
• Most Dodecad Project members belong to cluster #12 which corresponds to White Utahns in the reference populations
• #12 seems to be the main Iberian cluster

The following IDs appear as outliers:

DOD004 DOD029 DOD030 DOD032 DOD033 DOD034 DOD036 DOD047 DOD060 DOD072 DOD088 DOD119 DOD126 DOD128 DOD132 DOD156 DOD157 DOD168 DOD169 DOD175 DOD186 DOD196 DOD240 DOD245 DOD251

You can also see where you fall on the first 2 MDS dimensions. Your co-ordinates can be found in this spreadsheet.

Genetic structure in North-Central Europe with the Galore approach

I first posted this in the comments of my other blog, but it is worth a post of its own.

Here is the result of applying MCLUST to a group of Central-North European populations. The maximum number of 13 clusters is reached with 5 MDS dimensions retained:

Some clusters are population-specific (e.g., #7 for Finns, #10 for Lithuanians, #12 for Russians). Some clusters are semi-specific (e.g., #3 for Hungarians, #1 for French). Some populations are split into multiple clusters (e.g., Orcadians or Germans).

Here is a neighbor-joining tree of these 13 clusters based on the first 5 MDS dimensions:

12, 13: Russians

9, 10: Balto-Slavs

5, 3, 2, 4, 1: Northwest Europeans

11: A couple of Hungarians (*)

7, 8: Orcadians

6: Finns

I generally frown upon phylogenies for human groups, as I believe that human genetic variation is better represented as a network due to lateral gene flow. However, this tree gives an idea of the relationship between clusters.

(*) It's interesting that these 2 Hungarians are the same ones that showed an elevated "Altaic" component in my K=15 analysis.

Assyrians vs. Armenians

Assyrians and Armenians have invariably been assigned to the same cluster in Clusters Galore analysis thus far. It is very important to understand the following two axioms:

• Populations that are separated by clustering analysis have enough genetic differences between them to allow such separation
• Populations that are not separated may, or may not have enough genetic differences between them to allow separation

The second axiom is extremely important. Let's elaborate:

• The clustering algorithm may have its limitations
• The number of markers may be insufficient
• The number of individuals may be insufficient

In the Galore approach, I use MCLUST, as good a clustering algorithm as any. I also have samples of 8 Assyrians and 7 Armenians, and use about 177k SNPs, which are enough to squeeze out differences even between related populations. So, what gives? Are Assyrians and Armenians really genetically indistinguishable?

What constitutes a cluster? MCLUST can divide a dataset into as many clusters as you want, but it also chooses the number of clusters to optimize the Bayes Information Criterion. But, it's important to note that the BIC is not some god-given arbiter of what a cluster is; it is best viewed as a guide to choose a good number of clusters, and not as a guarantee that this is the true number of clusters that a population can be subdivided in.

Clustering Assyrians and Armenians, assuming 2 clusters

To that end, I decided to cluster Assyrians and Armenians, forcing MCLUST to infer 2 clusters. I only retained 2 MDS dimensions, as these are enough to distinguish between 2 groups. Here is the MDS plot:

We can observe that Assyrians can be distinguished from Armenians along Dimension 1, and there is also some structure in Assyrians, with 3 of them forming a mini-cluster on top, 4 of them forming another mini-cluster at the bottom, and 1 of them being closer to the Armenians.

By applying MCLUST with K=2, all 7 Armenians and 1 Assyrian are assigned to a cluster, and the remaining 7 Assyrians to another. Thus, the two groups can be separated from each other, although their differences (due to the factors I mentioned) are not large enough to lead to an improvement of the BIC.

With more individuals., it's possible that the BIC too will be able to track the improvement in likelihood that adding two clusters will produce.

And, indeed, it may be possible that some of these populations will be further subdivided. If more Assyrians join the project, for example, it may be the case that the two apparent Assyrian clusters will emerge, or it will be the case that the space between them will be filled by currently unsampled individuals.

As more individuals join the Project, ever-finer distinctions can be uncovered.

NB: I make the case for Assyrians and Armenians, but I have also tried this for other closely related populations who were assigned to the same cluster by Galore analysis, namely Spaniards and Portuguese. In that case, however, I was not able to find a clean solution with K=2. Once again, this may mean that Spaniards and Portuguese are not genetically that different, or it may mean that their difference will be revealed with more individuals joining the project.

Clusters galore, K=50 for Dodecad Project members (up to FFD055)

I am now repeating the Clusters galore analysis for Family Finder data (for more info, see the previous post on 23andMe data).

With 14 MDS dimensions retained, there were 50 clusters inferred in the optimal solution by MCLUST.

The results spreadsheet has rows for the 54 project participants in the first rows: each row is the probability that you belong to a particular cluster. This is followed by the reference populations where each row has the number of individuals (for that populations) that is assigned to a particular cluster.

There are also some outliers in this analysis:

FFD002 FFD004 FFD007 FFD012 FFD015 FFD016 FFD021 FFD022 FFD023 FFD038 FFD046

Check what an outlier is in the context of this analysis, and what it means.

Interestingly, because of the smaller number of Family Finder participants some previously defined clusters (for 23andMe data) such as the "Finnish" cluster do not appear here. This is not surprising at all, because for a cluster to be defined several individuals from that population must be present in the data.

Many continental Europeans of this type ended up in cluster #2. Some others, like FFD048 who is Lithuanian were assigned to the proper cluster #9, centered on Lithuanians.

This underscores the importance of having more people join the Project at the next available opportunity. This will not only create new clusters for individuals who are currently the only representatives of their populations, but it may also split already existing clusters if regional sub-populations are detected.

It is also important for project participants to drop a note at the ancestry thread, to help others make better sense of their results.

What's next for Clusters Galore analysis

The first few runs of the Clusters Galore analysis have proven quite successful; I've posted another one on the HGDP panel in my other blog.

Now, it is time to assess the results and see what improvements can be made. I see a few avenues for improvement:

Outliers

Clusters, by definition, are composed of at least 2 individuals. Individuals who are the only representatives of their populations (e.g. if a Pygmy or an Icelandic+Armenian mix) will, by necessity, attach themselves to the closest cluster (e.g., to Yoruba, or to some Central European population), even though they are not necessarily close to that population.

Outlier detection is a difficult problem, but I will try some ideas on how to tackle it.

Phantom clusters

mclust is resilient to phantom clusters, i.e., clusters of "misfits" who don't belong in any other populations but are banded together erroneously by the algorithm. That is inevitable in an automated procedure, especially one that is pushing the limits of ancestry inference. Phantom clusters are, by their nature, transient, so there are some ideas on how to avoid them and how to focus on very robust and repeatable clusters.

Typicality

Being part of a cluster tells you nothing about how "typical" a member of the cluster you are, i.e., how close to the average. This problem is exacerbated by the fact that the clusters inferred by mclust may have varying shape, size, and orientation.

Nonetheless there are ideas on how to quantify members' typicality, and I will explore them. Please note that typicality is not necessarily the same as "purity". For example, an elongated cluster of African Americans will have typical members with 20% European admixture, but the "purest" African Americans will have 0% European admixture and be very atypical of their group as a whole. Similarly, typical Turks have 5-6% East Eurasian admixture, but people with 10% East Eurasian admixture are less typical, but more likely to be descended from central Asian Turkic people.

Any new technique will have its birth pains, and hopefully myself and others will help identify them and resolve them.

Galore analysis improved, plus K=56 results of Clusters Galore analysis for Dodecad Project members (up to DOD236)

For background, please read the post on the K=48 analysis and links therein.

As I mentioned in the previous posts, my technique depends on the use of MDS to reduce the dimensionality of genomic data from 177,000 SNPs or so to a few dozen dimensions capturing most of the variance.

Subsequently, mclust a state of the art clustering algorithm is applied on the MDS representation: this iterates between choices of K, the number of clusters, trying clusters of different shape, volume, and orientation, and chooses the optimal clustering, maximizing the Bayes Information Criterion. In simpler terms, it finds as much detail as possible in the data but penalizes too ornate models and avoids finding "ghost" clusters that are not really supported.

These are clusters derived from data of unlabeled individuals. The only human input into the process is the number of MDS dimensions to retain.

In my previous K=48 analysis, I retained 30 dimensions, but I also noted that this is not really optimal. Choosing more, or less, dimensions might lead to even better resolution (higher K).

More dimensions = more possible ways to distinguish between individuals, but also, possibly, more noise, as individuals might not be "clustered" in them.

Fewer dimensions = less possible ways to distinguish between individuals, but also, possibly, less noise from the uninformative higher dimensions.

Thus, the question arises: how many dimensions to retain?

Here is a plot of the optimal number of clusters inferred, depending on how many dimensions I chose to retain:

As you can see, when a few dimensions are retained, relatively few clusters are inferred, while as the number of dimensions goes beyond a certain point, the number of clusters starts to decrease again, as more noise is added (*)

The number of clusters peaks (see figure) at 16 and 22 dimensions retained; both of these produce 56 different clusters in the optimal solution.

Here are the results for Dodecad Project members (up to DOD236) with K=56 and 16 dimensions retained. In comparison to the previous K=48 analysis, we are now able to:

1. Split CEU White Utahns (#1) from French (#15)
2. Split CEU White Utahns (#1) from continental Germanics (#14)
3. Split French (#15) from Spaniards (#2)
4. Split Armenians (#7) from Turks (#19)
5. Split Slavs (#23) from Balts (#26)
6. Split Cypriots (#30) from Sephardic Jews (#21)

The most astonishing finding is, however, at least for me, the emergence of a cluster (#16) comprised in the great majority by people from Greece and Southern Italy, with very few individuals from elsewhere. Notice that #16 has absolutely no representation in the reference populations, which lack South Italians and Greeks.

Once again, I urge participants to help themselves and others by leaving a comment in the ancestry thread.

(*) The change is not, however, smooth. The more general problem is to choose which dimensions to retain, rather than choosing how many of the first ones to retain. The first few dimensions of MDS capture a decreasing portion of variance, but the data are not guaranteed to be "split" in them. However, this is a much harder problem, as we have to figure out (i) how many dimensions to retain, and (ii) which ones. Even if we fix (i), by choosing to retain, e.g., 10 dimensions, we still have to choose which 10: this is close to half a billion different combinations of which 10 to choose from the total of 38 possible candidate ones.

Results of Clusters Galore analysis for Dodecad Project members (up to DOD236), K=48

This is the result of the new type of ancestry analysis I have recently devised. For background, please read:

• Clusters galore: extremely fine-scale ancestry inference
• Clusters galore with Dodecad populations
• Clusters galore: less is more, or, pushing the limits of ancestry inference

In total 894 individuals were included in this analysis, 202 Dodecad Project members and 692 from the published references. 30 MDS dimensions were retained, and mclust was run with a maximum number of clusters = 60. In the optimal solution, 48 clusters were inferred.

At the beginning of the results spreadsheet are the 202 Dodecad Project members and their probabilities of belonging to the 48 clusters. This is followed by the 36 reference populations, listing how many individuals from each one were assigned to each of the 48 clusters.

To help you interpret these results, you might want to consult the individual and population ADMIXTURE spreadsheets, as well as the information Project members have chosen to reveal about themselves. Feel free to add your own information in that thread.

Here are the results for some people who have chosen to reveal information about themselves:

• Spencer Wells is DOD162 and he has 28% probability of being in the CEU (White Utahn)/French cluster #1, and 72% probability of being in cluster #3 which is CEU (White Utahn) specific (in the reference populations) and in which 29 Project members are also assigned (almost all of them Northwestern Europeans)
• Razib of Gene Expression is DOD075 and is assigned in cluster #15, which includes Gujarati and North Kannada individuals
• pconroy, submitted three samples, DOD097 is Sicilian fall in cluster #9 to which many Cypriots and Sephardic Jews belong, and many Project members of South Italian/Sicilian background; DOD098 and DOD099 are his Irish parents: DOD098 is almost evenly split between #1 and #3 (like Wells), and DOD099 is 97% in #3
• Adriano Squecco DOD139 is North Italian and is in cluster #2, in which 25/25 reference Tuscans and 10/12 reference North Italians belong
• Lacko DOD083 is 100% southern Polish and he falls in cluster #20, in which all reference Belorussians and Lithuanians fall
• Mike Maddi DOD021 is Sicilian and is in cluster #9 like DOD097, showing some probability (13%) of also being in #2 like Adriano Squecco
• An Anonymous Pole falls in #20 like Lacko and the reference Slavs
• Ilmari DOD003 and Ari DOD131 are Finns and fall in cluster #13. This is an interesting one, as it does not occur at all in the reference populations; I'll let you guess what population it's centered on.
• Eastara's mother (DOD025) is Bulgarian and falls in cluster #10, which is centered in Romanians in the reference populations, but note that there are also 13 Dodecad Project members who fall in it, many of them from different parts of the Balkans. I hope more people from the Balkans will contact me for inclusion in the project, as I am sure that finer-scale can be achieved there with increased participation.
• Basar (DOD049) is half-Anatolian Turk and half-Laz. He falls in the cluster encompassing Armenians and Turks in the reference populations.
• Bubba (DOD066) is North German with a pinch of Danish and falls in cluster #3
• afpjr (DOD014) is half Greek and half Italian/Sicilian; he falls in cluster #9 (95%)
• Francesc (DOD217) is Catalan and falls in cluster #17, centered on Spaniards in the reference populations.

The Project Greeks and Mixed Greeks fall in clusters #2, 9, 10 tying them to Italy and the Balkans, as might be expected. I hope more Greeks will decide to participate in the Project, so we can discover more interesting patterns in our population.

I cannot stress enough how revealing non-identifying ancestry information in the relevant thread will help both yourselves and others make better sense of these and future results.

There are clusters composed entirely of Dodecad Project members (e.g., the aforementioned one), and others which are centered on one or two reference populations, but encompass a wider variety of non-represented populations. So, please take the time to leave a comment in the ancestry thread.

This is not the end of the story. There are more clusters to be discovered in the data; the inclusion of 200+ new samples in this analysis has caused new clusters to appear and distinctions that were previously detected to "fold back" (e.g., between Armenians and Turks). I am currently investigating how the choice of number of MDS dimensions to retain affects the number of inferred clusters in the optimal solution.