PCA projection bias in ancient DNA studies

Many Principal Component Analyses (PCA) in papers on ancient genomes clearly suffer from projection bias. However, most people don't seem to understand this problem and the impact it can have on the interpretation of the data.

Here's a demonstration of this effect using two PCA. In the first PCA, La Brana-1, a Mesolithic genome from Iberia, was projected onto the PC eigenvectors computed with modern individuals from the HGDP. However, in the second PCA the ancient genome was run together with these samples. Note the clear difference between the two outcomes.

The second outcome does look a bit strange, but it's actually the correct one, because it's now an established fact that Mesolithic hunter-gatherers, like La Brana-1, were clearly outside the range of modern European, and indeed West Eurasian, genetic variation.

For a technical discussion of this problem, which is also sometimes known as "shrinkage", refer to Lee et al. 2012.

The above experiment with La Brana-1 was run with PLINK 2, which is freely available here, using just over 16K SNPs. Only markers with a read depth of 4x or higher were considered, and the marker set was further pruned to account for no-calls (--geno 0.005), LD (--indep-pairwise 200 25 0.4), and minor allele frequency (--maf 0.05).

The really old Europe is mostly in Eastern Europe

A new version of the Lazaridis et al. ancient genomes preprint has just appeared at arXiv (see here). It includes several new Principal Component Analyses (PCA), TreeMix graphs, a ChromoPainter/fineSTRUCTURE co-ancestry matrix, and an updated ADMIXTURE analysis. The revised text underlines the relatively close genetic relationship between indigenous European hunter-gatherers and present-day Eastern Europeans:

The co-ancestry matrix (Fig. S19.3) confirms the ability of this method to meaningfully cluster individuals. We highlight two clusters: Stuttgart joins all Sardinian individuals in cluster A and Loschbour joins a cluster B that encompasses all Belarusian, Ukrainian, Mordovian, Russian, Estonian, Finnish, and Lithuanian individuals. These results confirm Sardinia as a refuge area where ancestry related to Early European Farmers has been best preserved, and also the greater persistence of WHG-related ancestry in present-day Eastern European populations. The latter finding suggests that West European Hunter-Gatherers (so-named because of the prevalence of Loschbour and La Braña) or populations related to them have contributed to the ancestry of present-day Eastern European groups. Additional research is needed to determine the distribution of WHG-related populations in ancient Europe.

Fig. S10.5 suggests that the main axis of differentiation in Europe when the subcontinent is considered as a whole may tend to Northeastern Europe rather than SSE/NNW (8). This is consistent with our analysis of ancestry proportions in European populations (Fig. 2B, Extended Data Table 3) which indicate a cline of reduced EEF (and increasing WHG) ancestry along that direction.

Citation...

Iosif Lazaridis, Nick Patterson, Alissa Mittnik, et al., Ancient human genomes suggest three ancestral populations for present-day Europeans, arXiv, April 2, 2014, arXiv:1312.6639v2

Another look at the Lazaridis et al. ancient genomes preprint

I've now had a chance to look over the Lazaridis et al. preprint a few times, and also take part in several online discussions about the results, at these blogs and elsewhere. So I thought it might be useful to put together another post on the paper to report what I've learned and reiterate a few points. First of all, to understand the results, it's really important to known what the four main ancestral components in this study represent:

- West European Hunter-Gatherer (WHG), based on an 8,000 year-old genome from Loschbour, Luxembourg

- Ancient North Eurasian (ANE), based on a 24,000 year-old genome from South Siberia (dubbed Mal'ta boy or MA-1)

- Early European Farmer (EEF), based on a 7,500 year-old genome from Stuttgart, Germany, belonging to the Neolithic Linearbandkeramik (LBK) culture

- Eastern non-African (ENA), this basically means East Eurasian, and is based on samples of present-day Onge, Han Chinese and Atayal from Taiwan

Now, from what I've seen online, many people seem to think that ANE is more East Asian than European, and can be considered a signal of pretty much any population expansion from the east into Europe. This is not true. ANE is Amerindian-like, but actually also very similar to WHG. In fact, they're equidistant from ENA:

The results of Table S12.1 provide suggestive evidence that Onge share more common ancestry with hunter-gatherers than with Stuttgart. All statistics involving two hunter-gatherer populations have |Z|<0.9, so ancient Eurasian hunter-gatherers are approximately symmetrically related to Onge, and they are all more closely related to them than is Stuttgart.

We next consider the relationship of ancient samples to East Asia using the set (Ami, Atayal, Han, Naxi, She). East Asians are more closely related to all hunter-gatherers than to Stuttgart, but there are no significant differences between hunter-gatherers (all such statistics have |Z|<1.1) (Table S12.2).

...

We have conveniently labeled MA1-related ancestry “Ancient North Eurasian” because of the provenance of MA1 in Siberia, but at present we cannot be sure whether this type of ancestry originated there or was a recent migrant from some western region.

The various Uralic, Turkic and Mongolian groups expanding into Europe, usually after the Bronze Age, no doubt carried significant ENA, so these groups can't be the source of the fairly high levels of ANE across Europe today, because most Europeans lack ENA. Below is a graph based on two f4 tests, comparing ANE and ENA ancestry among Europeans, this time with the Han Chinese as ENA proxies. Note that most of the samples fall within a cline that runs from the Stuttgart sample to Estonians. The only outliers in the direction of the Han are groups from current or former Uralic and Turkic speaking areas of Europe.

ANE was actually present in Scandinavia during the Mesolithic, because Motala12, the 8,000 year-old hunter-gatherer genome from Sweden, has an ANE ratio of 19%. But this isn't enough to explain the ANE levels carried by most present-day Europeans, so it's very likely there were at least two expansions of ANE into Europe.

Considering that Loschbour and Stuttgart totally lack ANE, it's plausible that a major wave of ANE moved across much of Europe sometime after the early Neolithic, but obviously before the Uralic and Turkic expansions, which, as per above, were rich in ENA. Based on recently published ancient mtDNA evidence from Central Europe (see here), Lazaridis et al. propose that this timeframe was the Copper and/or Bronze Age.

This of course is the generally accepted Proto-Indo-European timeframe. Indeed, the theory I put forward in the previous blog entry (see here) that most of the ANE in Europe today was the result of the Proto-Indo-European expansion, probably from Eastern Europe, looks even better on closer inspection.

Note the elongated cline formed by the European samples running from WHG to EEF on Fig 2B, shown below. It correlates well with latitude, and very likely reflects northward migrations of Neolithic farmers into Europe from the Mediterranean Basin, followed by isolation-by-distance. In other words, this cline probably took thousands of years to form.

On the other hand, there is no cline running from WHG/EEF to ANE, but all of the Indo-European and/or Eastern European samples are fairly evenly lifted up towards ANE relative to a few outliers. These outliers are all southwestern Europeans: Basques, Pais Vasco (Basque Country) Spaniards, southern French and Sardinians.

Of course, southwestern Europe is the most distant part of the continent from the generally accepted Indo-European homeland near the middle Volga. Moreover, Basques don't speak an Indo-European language, while Sardinians were only Indo-Europeanized during historic times.

Indeed, even though a couple of tables in the study report considerable ANE ancestry among Basques and Pais Vasco Spaniards, the authors admit that this need not be the case. For instance:

We next attempted to fit individual West Eurasian populations as a mixture of Loschbour and Stuttgart, as representatives of Early European farmers and West European Hunter Gatherers.

Fig. 1B suggests that this is not possible, as most Europeans form a cline that cannot be reconciled with such a mixture [Davidski's note: I think they actually mean Fig. 2B]. Nonetheless, for Sardinians (Extended Data Table 1), the most negative f3-statistic is of the form f3(Test; Loschbour, Stuttgart), which suggests that at least some Europeans may be consistent with having been formed by such a mixture. We thus fit each European population into the topology of Fig. S12.6. Only Basques, Pais_Vasco, and Sardinians, can be fit successfully with this model. Fig. S12.8 shows a successful fit.

Most European populations cannot be fit as this type of 2-way mixture and, intuitively, this is due to their tendency (Fig. 1B) towards Ancient North Eurasians that is not modeled by such a mixture.

Another intriguing thing about the results shown in Fig 2B is that the expansions of ANE across Europe appear not to have disturbed the presumably Neolithic WHG/EEF cline to any great extent. What this suggests is that ANE was spread largely independently of EEF and even WHG. In other words, the groups that pushed ANE deep into Europe probably had very high ratios of this component. This also seems to be true for the groups that brought ANE to the Near East:

A geographically parsimonious hypothesis would be that a major component of present-day European ancestry was formed in eastern Europe or western Siberia where western and eastern hunter-gatherer groups could plausibly have intermixed. Motala12 has an estimated WHG/(WHG+ANE) ratio of 81% (S12.7), higher than that estimated for the population contributing to modern Europeans (Fig. S12.14). Motala and Mal’ta are separated by 5,000km in space and about 17 thousand years in time, leaving ample room for a genetically intermediate population. The lack of WHG ancestry in the Near East (Extended Data Fig. 6, Fig. 1B) together with the presence of ANE ancestry there (Table S12.12) suggests that the population who contributed ANE ancestry there may have lacked substantial amounts of WHG ancestry, and thus have a much lower (or even zero) WHG/(WHG+ANE) ratio.

So perhaps the 17,000 year-old Afontova Gora 2 (AG2) genome from Central Siberia, classified as part of the ANE meta-population by Lazaridis et al., is genetically the closest sample we have to the Proto-Indo-Europeans? Based on a couple of the PCA from Lazaridis et al. (below) and Raghavan et al. (see here), this genome doesn't appear to be 100% ANE. My very rough estimate is 85/15 ANE/WHG.

If my assumptions are correct here, then it's no wonder that this Bronze Age Danish sample (M4) from the recent Carpenter et al. paper (see here) shows a clear shift towards the Americans on the global PCA. M4 is better known as "the old man" from the giant Borum Eshøj barrow (see here), presumably built by some of the earliest Indo-Europeans in Scandinavia. We can probably expect such Afontova Gora 2-like results from many European samples archeologically linked to the early Indo-Europeans.

Ancient human genomes suggest (more than) three ancestral populations for present-day Europeans

This new preprint at bioRxiv is quite the Christmas present for those of us with a passion for European genetics and prehistory. It's the first paper to report on full genomes from Mesolithic and Neolithic Europe.

All of the successfully tested Mesolithic Y-chromosomes, one from Luxembourg and four from Motala, Sweden, belonged to haplogroup I. This probably won't come as a surprise to many people, as this marker was always the main candidate for Europe's indigenous Y-haplogroup. However, three of the results fell into haplogroup I2a1b, and none into I1, which is today the most common Y-haplogroup in most of Scandinavia.

What this suggests is that I1 expanded after the Mesolithic and replaced most of the I2a1b across Northwestern Europe. I'd say these were mostly expansions from North-Central Europe, although recent chatter on the web suggests that two distinct I1 lineages might have arrived in North-Central Europe from Eastern Europe at different times.

All of the Mesolithic mtDNA sequences belonged to haplgroups U2 and U5, which is line with past results. The single Neolithic sample, from a 7500 year-old Linearbandkeramik (LBK) site in Stuttgart, Germany, belonged to mtDNA haplogroup T2. Again, not very surprising considering what we've seen to date.

The genome-wide results, on the other hand, are not as straightforward. The basic upshot is that Northern Europeans are mostly of indigenous European hunter-gatherer origin, while Southern Europeans are largely derived from Neolithic farmers of mixed European and Near Eastern origin. But the authors identify a minimum of three ancestral populations from their stats (WHG, EEF and ANE), and four meta-populations from the available ancient data (WHG, EEF, ANE and SHG). Here are brief summaries of each of these groups:

West European Hunter-Gatherer (WHG): this ancestral component is based on an 8,000 year-old forager from the Loschbour rock shelter in Luxembourg (one of the individuals mentioned above belonging to I2a1b). The WHG meta-population includes the Loschbour sample and two Mesolithic individuals from the La Brana Cave in Spain. However, today the WHG component peaks among Estonians and Lithuanians, in the East Baltic region, at almost 50%.

Early European Farmer (EEF): apparently this is a hybrid component, the result of mixture between "Basal Eurasians" and a WHG-like population possibly from the Balkans. It's based on the aforementioned LBK farmer from Stuttgart, but today peaks at just over 80% among Sardinians. Apart from the Stuttgart sample, the EEF meta-population includes Oetzi the Iceman and a Neolithic Funnelbeaker farmer from Sweden.

Ancient North Eurasian (ANE): this is the twist in the tale, a component based on a previously reported genome of a 24,000 year-old Upper Paleolithic forager from South Central Siberia, belonging to Y-hg R*, and known as Mal'ta boy or MA-1 (see here). This component was very likely present in Southern Scandinavia since at least the Mesolithic (see the summary of SHG below), but only seems to have reached Western Europe after the Neolithic. At some point it also spread into the Americas. In Europe today it peaks among Estonians at just over 18%, and, intriguingly, reaches a similar level among Scots. However, numbers weren't given for Finns, Russians and Mordovians, who, according to one of the maps, also carry very high ANE, but their results are confounded by more recent Siberian admixture (see the discussion on the European outliers below). The ANE meta-population includes Mal'ta boy as well as a late Upper Paleolithic sample from Central Siberia, dubbed Afontova Gora-2 (AG2).

Scandinavian Hunter-Gatherer (SHG): this is a meta-population made up of Swedish Mesolithic and Neolithic forager samples from Motala and Gotland, respectively. It's a more easterly variant of WHG, with probable ANE admixture.

Below are the two most important figures from the paper: a) the three-way mixture model that is a statistical fit to the data, and b) a plot of the proportions of ancestry from each of the three inferred ancestral populations. As per above, East Baltic populations are the most WHG, which is somewhat curious, because they mostly carry Y-DNA R1a and N1c1.

So if not for the ANE, we'd simply have a two-way mixture model between indigenous European foragers and migrant Near Eastern farmers, at least for most Europeans anyway. Moreover, the seemingly late and sudden arrival of ANE in much of Europe is important, because it's a smoking gun for a major population upheaval across the continent during the Late Neolithic/Early Bronze Age.

Interestingly, archeological data suggest that this was also the period which saw the introduction of new social organization and perhaps Indo-European languages across most of Europe. None of this was lost on the authors of the paper, but it appears they'd rather be cautious pending more ancient genomic data, because they chose not to explicitly mention the Indo-Europeans.

This study raises two questions that are important to address in future research. A first is where the EEF picked up their WHG ancestry. Southeastern Europe is a candidate as it lies along the geographic path from Anatolia into central Europe, and hence it should be a priority to study ancient samples from this region. A second question is when and where ANE ancestors admixed with the ancestors of most present-day Europeans. Based on discontinuity in mtDNA haplogroup frequencies in Central Europe, this may have occurred during the Late Neolithic or early Bronze Age ~5,500-4,000 years ago35. A central aim for future work should be to collect transects of ancient Europeans through time and space to illuminate the history of these transformations.

...

The absence of Y-haplogroup R1b in our two sample locations is striking given that it is, at present, the major west European lineage. Importantly, however, it has not yet been found in ancient European contexts prior to a Bell Beaker burial from Germany (2,800-2,000BC)12, while the related R1a lineage has a first known occurrence in a Corded Ware burial also from Germany (2,600BC)13. This casts doubt on early suggestions associating these haplogroups with Paleolithic Europeans14, and is more consistent with their Neolithic entry into Europe at least in the case of R1b15, 16. More research is needed to document the time and place of their earliest occurrence in Europe. Interestingly, the Mal’ta boy belonged to haplogroup R* and we tentatively suggest that some haplogroup R bearers may be responsible for the wider dissemination of Ancient North Eurasian ancestry into Europe, as their haplogroup Q relatives may have plausibly done into the Americas17.

No doubt, a lot of people will now be wondering about the main source of the ANE that apparently rushed into Europe at the onset of the metal ages. The Siberian steppe will probably be the favored option for many, since this is where Mal'ta boy and Afontova Gora-2 were dug up. However, I'm pretty sure the source was Eastern Europe.

First of all, as already mentioned, it seems that ANE was present in Sweden during the Mesolithic (Figure S12.7 shows around 19% ANE in the Motala12 sample). Secondly, despite the ANE and WHG being classified as separate ancestral and meta-populations, the differences between them appear to be clinal rather than discrete, which I think can be seen in the PCA and ADMIXTURE results from the study (see here and here). Thus, I'd expect a lot more ANE in Eastern Europe during the Mesolithic than in Scandinavia. Thirdly, it's likely that the ancestors of modern Uralic speakers were in Siberia very early, possibly during the Mesolithic, and they were probably East Eurasians aka. Eastern non-Africans (ENA), which ANE is not.

Indeed, latest linguistics research suggests that the pre-proto-Uralics migrated at some point from Siberia into the southern Urals, in far eastern Europe. The Uralics proper then expanded from the southern Urals, probably during the Bronze Age, both to the east and west, as far as the Baltic. This Uralic expansion is certainly reflected in the Lazaridis et al. data, and it's not the only relatively late migration into Europe that shows up in their stats.

While our three-way mixture model fits the data for most European populations, two sets of populations are poor fits. First, Sicilians, Maltese, and Ashkenazi Jews have EEF estimates beyond the 0-100% interval (SI13) and they cannot be jointly fit with other Europeans (SI12). These populations may have more Near Eastern ancestry than can be explained via EEF admixture (SI13), an inference that is also suggested by the fact that they fall in the gap between European and Near Eastern populations in the PCA of Fig. 1B. Second, we observe that Finns, Mordovians, Russians, Chuvash, and Saami from northeastern Europe do not fit our model (SI12; Extended Data Table 3). To better understand this, for each West Eurasian population in turn we plotted f4(X, Bedouin2; Han, Mbuti) against f4(X, Bedouin2; MA1, Mbuti), using statistics that measure the degree of a European population’s allele sharing with Han Chinese or MA1 (Extended Data Fig. 7). Europeans fall along a line of slope >1 in the plot of these two statistics. However, northeastern Europeans fall away from this line in the direction of Han. This is consistent with Siberian gene flow into some northeastern Europeans after the initial ANE admixture, and may be related to the fact that Y-chromosome haplogroup N 30, 31 is shared between Siberian and northeastern Europeans32, 33 but not with western Europeans. There may in fact be multiple layers of Siberian gene flow into northeastern Europe after the initial ANE gene flow, as our analyses reported in SI 12 show that some Mordovians, Russians and Chuvash have Siberian-related admixture that is significantly more recent than that in Finns (SI12).

The authors are actually referring to the Kargopol Russians from the HGDP in that quote. But from my own analyses with a wide variety of samples from Russia, I know that other Russians show similar levels of Siberian admixture to Belorussians, Ukrainians and Estonians.

In any case, this of course means that there are more than three ancestral populations for present-day Europeans, albeit not all of them influenced all Europeans. Also, it's very clear that to learn all the details about the peopling of Europe, these sorts of studies really need to start focusing on the large swath of land that stretches from present-day Poland to the Urals. In other words, Eastern Europe.

I was also going to discuss the genetically inferred pigmentation of the ancient individuals, but, because of the small sample size, there's not much to discuss at this stage. The Loschbour forager possibly had blue eyes (50% chance), but dark hair and skin. On the other hand, the Stuttgart farmer definitely had dark eyes and hair, but relatively light skin. I wonder if this swarthy hunter-gatherer skin complexion has anything to do with the fact that today lots of people from around the Baltic tan really well?

Citation...

Iosif Lazaridis, Nick Patterson, Alissa Mittnik, et al., Ancient human genomes suggest three ancestral populations for present-day Europeans, bioRxiv, Posted December 23, 2013, doi: 10.1101/001552

See also...

Another look at the Lazaridis et al. ancient genomes preprint

The really old Europe is mostly in Eastern Europe

EEF-WHG-ANE test for Europeans

First genome of an Upper Paleolithic human

ADMIXTURE analysis of Allentoft et al. and Haak et al. ancient genomes

West Eurasian cluster analysis: 13 clusters from 18 dimensions

I ran a quick Mclust analysis to get a better idea of the substructures in my recently updated dataset of West Eurasian samples. Mclust found that the optimal outcome was produced with 18 dimensions of genetic variation and 13 clusters, the latter of which are superimposed on a two dimensional MDS plot below. I chose the labels for the clusters myself and flipped the canvass to fit geography.

Here you can see the 13 clusters superimposed on all possible combinations of the 18 dimensions. Clicking on the image will take you to a 10.3MB PDF file.

It's interesting to note the presence of the very tight Jewish cluster, which includes Ashkenazi, Sephardic and Moroccan Jews. The Basques and Sardinians also cluster together, despite being clearly distinct from each other in the fist two dimensions. This is fascinating because these two groups have been mentioned a few times now in various studies and presentations as being the best modern proxies for Europe's Neolithic farmers.

The widespread Central and Eastern European cluster mostly includes individuals from populations that aren't easily characterized in these sorts of tests, and that's basically because they're of mixed origin. Indeed, I suspect things would look somewhat different in that part of the plot if I had more sizable numbers from Germany, Scandinavia, Poland and nearby areas.

Mclust can produce many more clusters than just 13 from the same data, but as per above, I wanted to see what would happen if it was asked to come up with the optimal solution. For more on this type of analysis check out the articles here, here and here.


Update 17/12/2013: On a related note, here's an Mclust analysis of West, Central and South Asia. The optimal result was obtained with 10 dimensions and 14 clusters. Please note that although some of the clusters have the same names as in the analysis above, they aren't the same clusters.

See also...

Eurogenes' North Euro clusters - phase 2, final results

Tracing the Indo-Europeans

As far as I can tell, these videos came online only a few weeks ago. They're from a conference titled "Tracing the Indo-Europeans: Origin and migration", which was held in Copenhagen late last year as part of the Roots of Europe project. I've had a quick look at the selection below, and the impression I get is that the guest speakers would rather eat shards of glass than accept that the Indo-European homeland was in Anatolia (ie. the so called Anatolian hypothesis). Also, interestingly, in the last video, Kristian Kristiansen discusses the possibility that the origin of the Maritime Bell Beaker culture was in the Aegean region, and that it might have been Proto-Celtic. If the sound is too low, use VLC Media Player and crank up the volume.

Adam Hyllested: Indo-European homeland and dispersals: Contemporary linguistic evidence

Guus Kroonen: The linguistic heritage of the European Neolithic: Non-Indo-European words in Germanic

David Anthony: Early Indo-European migrations, economies, and phylogenies

Kristian Kristiansen: Trade, travels and the transmission of cultural change in the Bronze Age

Morten Allentoft: Using ancient DNA to study human evolution and migration

David Anthony: Wheeled vehicles, horses, and Indo-European origins

Kristian Kristiansen: The Bronze Age expansion of Indo-European languages

South Asian R1a in the 1000 Genomes Project

After a recent update, the 1000 Genomes project now includes 62 individuals of South Asian origin belonging to Y-DNA haplogroup R1a-M17. Their full Y-chromosome sequences have been analyzed by Semargl and Maximus (aka. YFull project), with some interesting but not unexpected results:

- All individuals belong to R1a-Z93, which appears to totally dominate South Asian R1a-M17.

- A single Punjabi from Lahore, northeastern Pakistan, is ancestral for the Z94 mutation, which is just below Z93. All the other individuals are derived for Z94.

- Six individuals - of Punjabi, Bangladeshi and Gujarati origin - are ancestral for L657 and Z2124, the two main mutations immediately below Z94.

- All individuals of South Indian and Sri Lankan origin are derived for L657 or Z2124.

- Based on this sample, there appears to be no substructure along ethnic or geographic lines within South Asian R1a-M17 derived for L657 and Z2124.

Thus, it seems the SNP diversity of South Asian R1a-M17 is low, and decreases from Pakistan, North India and Bangladesh to South India and Sri Lanka. In comparison, there are only 12 European R1a individuals in the 1000 Genomes sample, and they represent all the major subclades of this haplogroup: R1a-Z283, R1a-Z93 and R1a-L664. Therefore, sampling bias can't be used as an argument for the more diverse result from Europe.

The lack of substructure along ethnic and geographic lines within South Asian R1a-L657 and R1a-Z2124 looks unusual, especially considering the caste system in India, and needs to be verified with more extensive sampling. However, if this outcome holds up, it'll suggest that paternal gene flow across South Asia has not been restricted by the caste system or geography. Then again, it could mean the caste system appeared after R1a-L657 and R1a-Z2124 arrived in South India via massive population movements from the north.

Below are all the results in as much detail as the current R1a SNP tree allows. Key: BEB - Bengali from Bangladesh; GIH - Gujaratai from Houston, Texas; ITU - Indian Telugu from the UK; PJL - Punjabi from Lahore, Pakistan; STU - Sri Lankan Tamil from the UK.

Z93+ Z94-
PJL - 1

Z94+ L657- Z2124- Z96-
BEB - 2 PJL - 3 GIH - 1

L657+,Y2+ etc.
1) Y9 (inc. Y7)
GIH - 7
STU - 4
ITU - 4
PJL - 8
BEB - 2

2) Y4+, Y8+, Y28+ (inc. Y6+)
GIH - 6
ITU - 6
PJL - 2
STU - 6
BEB - 5

Z2125+ (Z2124+ Z2122- Z2123-)
PJL - 1

Z2123+ (Z2124+ Z2122-, Z2125-)
PJL - 2
STU - 3
BEB - 1
ITU - 6
GIH - 2

So who's the most (indigenous) European of us all?

Basically, the first map below reveals the answer. It shows the spread of a European specific cluster from a global-wide ADMIXTURE analysis at K=8 (eight ancestral populations assumed), which I call "North European". Thus, genetically, the most European populations are found around the Baltic Sea, and in particular in the East Baltic region. In my genome collection, samples from Lithuania clearly and consistently score the highest percentages in ADMIXTURE clusters specific to Europe. However, I suspect that if I had Latvians with no known foreign ancestry going back more than four generations, they'd come out the "most European". Hopefully we can test that in the near future.

Below are the fifteen Eurogenes sample sets that scored the highest levels of membership in the North European cluster. The list only includes groups with five or more individuals present in the analysis, so some populations, like Estonians or Danes, weren't included, even though they easily made the cut. The spreadsheet with all the results from this run can be seen here. A table of Fst (genetic) distances between the eight clusters is available here.

Lithuanians 77%
Finns 74%
Belorussians 70%
Swedes 69%
Norwegians 68%
Kargopol Russians 68%
Russians 68%
Poles 68%
Erzya 66%
Ukrainians 66%
Moksha 66%
Orcadians 63%
HapMap Utah Americans (CEU) 63%
Irish 63%
British 62%

So why did I pick the results from K=8, and not some other K, like 2, 10, or 25? Well, it's not possible to evaluate who is more European without a European-specific cluster (ie. modal in Europeans, with a low frequency outside of Europe). Provided that a decent number and range of global and West Eurasian samples are used in the analysis, such clusters begin appearing at around K=5 or K=6, and start breaking up into local clusters from about K=9. I found that runs below K=8 produced European clusters that spilled too generously outside of the borders of Europe. On the other hand, runs above K=8 produced European clusters that weren't representative of enough European groups (ie. too localized). But the European cluster from K=8 was pretty much perfect, and I think that's obvious from the map. In fact, I can hardly believe how well it fits the modern geographic concept of Europe - north of the Mediterranean and west of the Urals. Amazing stuff.

There are two other clusters that show up across Europe in non-trivial amounts - Mediterranean and Caucasus (see maps below). These can also be thought of as native European clusters, since they've been on the continent for thousands of years. However, their peak frequencies are found in West Asia, so they're not particularly useful signals of European-specific ancestry.

So what do these three clusters show exactly? They represent certain allele frequencies in modern populations, and in fact, these can change fairly rapidly due to admixture, selection, and genetic drift. So claiming that such clusters represent pure ancient populations is unlikely to be true in most cases, if ever. However, I don't think there's anything wrong in saying that, when robust enough, they can be thought of as signals of ancestry from relatively distinct ancestral groups.

Indeed, anyone who's read up on the prehistory of Europe, knows that there are three general Neolithic archeological waves to consider when trying to untangle the story of the peopling of Europe. These are Mediterranean Neolithic, Anatolian Neolithic and Forest Neolithic (for example, see here).

Mediterranean Neolithic refers to a series of migrations from West Asia via the Mediterranean and its coasts. The areas most profoundly affected by these movements include the islands of Sardinia and Corsica, and the Southwest European mainland. Anatolian Neolithic describes migrations into Europe from modern day Turkey, mostly into the Balkans, but also as far as Germany and France. At the moment, Forest Neolithic of Northeastern Europe is something of a mystery. However, the general opinion is that it was largely the result of native Mesolithic hunter-gatherers adopting agriculture.

Obviously, it's very difficult to dismiss the correlations between these three broad archeological groups and the European and two European/West Asian clusters produced in my K=8 ADMIXTURE analysis. Is it a coincidence that the Mediterranean cluster today peaks in Sardinia, which has been largely shielded from foreign admixture since the Neolithic, and today forms a very distinct Southern European isolate? Why does the North European cluster show the highest peaks in classic Forest Neolithic territory? And why does the Caucasus cluster radiate in Europe from the southeast, which is where Anatolian farmers had the greatest impact? These can't all be coincidences, and I'm willing to bet that none of them are. I'm convinced that the three clusters from my K=8 run are strong signals from the Neolithic, and the North European cluster also from the Mesolithic.

Eventually, these issues will be settled with ancient DNA data, in a much more comprehensive way than ever possible using modern genomes. We've already seen some preliminary results, mostly from Mesolithic, Neolithic and Bronze Age sites around Europe, so perhaps it's useful to ask whether my ADMIXTURE analysis and commentary here mirror these early findings? I think they do. For instance, here's an interesting conclusion regarding the East Baltic area from a study on ancient Scandinavian mtDNA by Malmström et al.

Through analysis of DNA extracted from ancient Scandinavian human remains, we show that people of the Pitted Ware culture were not the direct ancestors of modern Scandinavians (including the Saami people of northern Scandinavia) but are more closely related to contemporary populations of the eastern Baltic region. Our findings support hypotheses arising from archaeological analyses that propose a Neolithic or post-Neolithic population replacement in Scandinavia [7]. Furthermore, our data are consistent with the view that the eastern Baltic represents a genetic refugia for some of the European hunter-gatherer populations.

I suppose there will be people wondering why I didn't take Sub-Saharan African, East Asian, and South Asian admixtures into account in my analysis. The reason is that I wasn't looking at which group was most West Eurasian, or Caucasoid. Based on everything I've seen to date, in my own work as well as elsewhere, the most West Eurasian group would probably be the French Basques from the HGDP. However, the differences between them, and certain groups from Northeastern Europe, like Northern Poles and Lithuanians, really wouldn't be that great anyway. I might do a write up about that at some point.


Credits...

- Maps by Eurogenes project member FR7

- Additional stats by Eurogenes project member DESEUK1


References...

Helena Malmström et al., Ancient DNA Reveals Lack of Continuity between Neolithic Hunter-Gatherers and Contemporary Scandinavians, Current Biology, 24 September 2009, doi:10.1016/j.cub.2009.09.017

Noreen von Cramon-Taubadel and Ron Pinhasi, Craniometric data support a mosaic model of demic and cultural Neolithic diffusion to outlying regions of Europe, Proc. R. Soc. B published online 23 February 2011, doi: 10.1098/rspb.2010.2678

Eurogenes' North Euro clusters - phase 2, final results

This is a continuation of my ChromoPainter analysis of Europeans from north of the Pyrenees, Alps and Balkans (see here). To obtain the most accurate results possible on my laptop, I increased the burn-ins and iterations in fineSTRUCTURE to 500K each (5 hour run in all, which is all I'm willing to put this machine through). The end product looks very similar to my initial analysis, in which I explored the data at 200K burn-ins and iterations. What I think this shows is that the results are robust, and I doubt they'd change much even after a couple of days of running fineSTRUCTURE.

Indeed, as mentioned in my previous blog entry, this appears to be the most detailed and accurate cluster analysis of this part of Europe produced anywhere to date. There are 21 clusters in all, with at least 20 looking like strong signals of genetic substructures across North, West, Central and East Europe (see spreadsheet for individual classifications). They include:

pop0 - West Finnish1: This is a pair of reference individuals, most likely from Western Finland, judging by their PCA and ADMIXTURE results. They are either from the same community, or have a very similar mix of very specific ancestries.

pop1 - Erzya + Moksha: This includes all of the Erzya and Moksha in the project, plus a Russian with recent Erzya ancestry. It's closely related to ethnic Russian clusters that stretch from Northwest Russia to near the Volga, and also to the Estonian cluster.

pop2 - South/Central Finnish: This is the largest Finnish cluster, and that's probably more than just the result of sampling bias. I would say that the greater part of the Finnish population would belong to this type of cluster, which occupies regions of highest population density within the country.

pop3 - Fenno-Scandian: This cluster includes a Northern Swede, a Swede with probable recent Finnish ancestry, and Finns with probable recent Swedish influence. I have a feeling that Finland Swedes and Aland Islanders would also be placed here more often than not.

pop4 - Northwest Russian/Southeast Finnish: Although this cluster includes only two individuals, it's definitely much more than just the result of two relatively closely related samples being in the same run. I'd hazard a guess that Northwest Russians with, say, significant Ingrian ancestry, would land here, and so would Finns with recent Russian ancestry.

pop5 - West Finnish2: Based on PCA and ADMIXTURE results, most of these Finns likely come from Western Finland, probably from places like Southern Ostrobothnia. They possibly also have some Swedish influence.

pop6 - West German: This cluster is based on individuals from Western and Northwestern Germany. It also includes a Dutchman, Austrian and people of mixed origin, like a Dane with French and German ancestry, and Americans with British, German, Scandinavian and/or Polish ancestry. In other words, this is where Northwestern Europe meets Central Europe.

pop7 - Vologda Russian: Most of the Vologda Russians from the HGDP land here, so this appears to be a local cluster. Judging from its phylogeny, it looks like a mix of North Slavic, Baltic and Finnic influences.

pop8 - East Finnish: All the project and reference Finns with substantial ancestry from new settlement areas of Eastern Finland appear in this cluster. No wonder then, that this is the cluster with the highest chunk count in this analysis.

pop9 - Estonian: This is a mixed cluster, including individuals from Estonia, and, as far as I know, Russians with substantial ancestry from near Estonia. As mentioned above, it's closely related to the Erzya + Moksha, Northwest Russian and Vologda clusters. However, it's clearly much more western than any of these clusters (for instance, see the PCA below), which suggests Germanic influence in its makeup.

pop10 - Cornish: Almost all of my Cornish samples from the 1000 Genomes Project feature in this very local cluster, which shows the highest chunk count among the Western European samples. The overall results suggest a lack of outbreeding in recent times.

pop11 - French/Belgian: Interestingly, this cluster includes the bulk of the French samples, a French Canadian, and two Belgians. On the other hand, the most northerly French are placed in the more cosmopolitan Northwest European cluster (see below).

pop12 - Lithuanian: All of the more or less pure Lithuanians fall in this cluster. Those that don't are a reference sample from Behar et al. 2009, who always appears very Belorussian like in other analyses, and here sits in the East Slavic cluster, and a project member with recent German ancestry (LIT3). The Western European influence carried by the latter pushes him into the Polish/West Ukrainian cluster, despite not having any documented Polish or Ukrainian ancestry.

pop13 - Northwest Russian: This cluster appears to be made up of Russians who have more Finnic, and/or perhaps Eastern Baltic, ancestry than the individuals in the East Slavic cluster. In other words, it's more northerly, less westerly, and more closely related to the Finnic-speaking Erzya, Moksha and Estonians.

pop14 - Irish + West British: Most Irish individuals fall in this cluster, as well as British samples from Western Scotland and Wales. It's tempting to correlate this cluster with Celtic genetic ancestry in the Isles.

pop15 - South/West Scandinavian: This is basically a Norwegian and Southern Swedish cluster. It also features Swedes from other parts of the country who most likely have some German, Walloon and/or French influence.

pop16 - East German: This cluster includes individuals with significant or even overwhelming Germanic ancestry, but also with very clear Western Slavic input. One of the individuals here is of mixed Polish, German and Swedish ancestry, which pretty much sums up the character of this cluster in a modern context. The presence of two Hungarians from Behar et al. 2009. isn't surprising, because Hungary was settled by both Germanic and Western Slavic groups from the early Middle Ages until modern times.

pop17 - Northwest European: I had reasonable hopes of breaking up this large cluster into a couple of units at least. However, that did not happen, and I don't think it will unless I obtain more samples from the relevant areas of Europe, like Holland and specific parts of the UK. I think the main reason this cluster failed to budge was because of its cosmopolitan nature. In other words, the samples here include some of the most outbred in the analysis, and this, coupled with the fact that they carry very similar ancestral components, means that fineSTRUCTURE doesn't have anything to latch onto to create divisions.

pop18 - East Scandinavian: This could also be called a Swedish cluster. It's almost entirely made up of Swedes, usually from Eastern or Southeastern Sweden, and/or occasionally with recent Finnish influence.

pop19 - Polish/West Ukrainian: The vast majority of the Poles fall in this cluster, and about half of the Ukrainians from Yunusbayev et al. 2011. Most of these Ukrainians appear to be from the Lviv district in the west, and some might even have fairly recent Polish and/or German ancestry. In fact, I would say the latter is a good bet for UkrLv240Y, who shows large Western European segments on several chromosomes.

pop20 - East Slavic: All of the Belorussians cluster here, and so do Russians from near Belorussia and Ukraine, and almost half of the Ukrainians from Yunusbayev et al. 2011 (those who show more easterly genetic characteristics). An individual of mixed Polish and Lithuanian ancestry also makes an appearance here, suggesting that one of the main factors differentiating this cluster from the Polish/West Ukrainian group is a higher level of Baltic admixture in the former.

pop21 - East Central European: This cluster is based on most of the Hungarians in my dataset, but it also includes a number of Western and Southern Slavs, often with significant German ancestry. Not surprisingly, this cluster shows very high affinity with both the East German and Polish/West Ukrainian clusters.

Let's now move on to some graphics. Below, in order of appearance, are the following: raw data coancestry matrix, showing the placement of individual samples; aggregate coancestry matrix, showing the populations (or clusters) described above; pairwise coincidence matrix, which is useful for spotting very recent ancestral ties; a PCA plot of the 21 clusters. More detailed ChromoPainter/fineSTRUCTURE PCAs of Western Europe can be found at this link.





Finally, those of you who wish to run your own experiments with the ChromoPainter datasheets from this analysis can download them here. Please note, the sheets don't reveal any raw or traits/disease data.

Eurogenes' North Euro clusters - phase 1, exploring the data

I have some preliminary results from a new intra-North Euro cluster analysis, using a cutting edge tool called ChromoPainter. More than 400 samples and 270K SNPs were tested, in linkage mode, and then the output processed in fineSTRUCTURE at 200K burn-ins and iterations. Like I say, the results should be treated as preliminary, but they already look better than any other cluster analysis I've ever seen dealing with Europe north of the Alps, Pyrenees and Balkans. The algorithm identified 21 clusters, with most located in Eastern and Northeastern Europe (see spreadsheet for details). Below are two plots showing how the clusters relate to each other via a tree diagram and heat maps – the first shows an aggregate view, and the second the individual samples.





It's interesting that the Baltic Finns seem to create clusters at a drop of a hat, but they also share the highest number of chunks, and the longest chunks, than any other group. Indeed, all of the Finnish clusters are closely related, and many of the individuals, especially from East Finland, even look like distant relatives on the heat map (note the ultra-hot, blue squares). On the other hand, the large Northwestern European cluster, featuring samples from across the UK, as well as from several nearby countries, is holding firm, and might be tough to break up in this analysis.

I have some theories about the reasons for the obvious genetic homogeneity and diversity in Western Europe, and these include the effects of the Black Death. It decimated many populations in the western half of the continent, thus encouraging migrations into emptied areas, and eventually leading to more open, mobile societies. It's an interesting subject, and I might write much more on it in the future. Meantime, here's a PCA plot from the ChromoPainter chunk counts data. Note the large distances spanned by groups from Northern and Eastern Europe, and the tight bundle of samples from the west, mostly from the UK, Ireland, France and the Low Countries. Interestingly, and perhaps counter-intuitively, it's the closely related Finns who take up most of the space on the plot.



The first component picked up by this PCA appears to be an Atlantic one. It peaks in the Cornish samples, but shows similar levels in all the British, Irish, French, Dutch and Belgians (post-Black Death mobility?). If we are to assume that I identified the component correctly, then it appears as if the East Finns, Vologda Russians, Erzya from the Middle Volga, and Lithuanians are the least “Atlantic” samples in this analysis. These groups, especially the East Finns, also happen to act like relative genetic isolates in many of my experiments (such as ADMIXTURE and MDS analyses). Thus, it seems they've been sheltered from significant gene flow from outside in recent times, including from the west, like German emigration to East Central Europe and Scandinavian influence in Western and Southwestern Finland.