skip to main |
skip to sidebar
Here I ran my white American project members against sample sets from the UK, Ireland, Scandinavia, France, Poland, Germany, South and West Finland, and others. I chose these reference samples to help flesh out potential signals of ancestry from different parts of North, Central and Eastern Europe. On the other hand, I left out other references from the north, like the Lithuanians, who tend to polarize Northern Europe into east and west, and contract the western cluster. I don't know what they do this, because from what I've just seen, they're really not very different from Northwest Europeans at haploblock level (results of that analysis to be published here soon). Perhaps it's a result of some freaky allele frequencies due to founder effect? Anyway, refer to the "Test samples" sheet to see which MDS you're on, and follow the instructions below...- download Gnuplot
- unpack Gnuplot
- download the MDS data NEU_MDS.zip
- place the relevant mds.dat into the "binary" Gnuplot folder
- type: splot 'mds1.dat' using 3:4:5:1 with labels
- spin the plot around and find some informative angles
To get large png files of the angles you like, say 2000x1500 pixels, try...
- type: set term png size 2000, 1500
- type: set output "mds1.png"
- type: splot 'mds1.dat' using 3:4:5:1 with labels
- download Gnuplot
- unpack Gnuplot
- download the MDS data SAS.dat
- place SAS.dat into the "binary" Gnuplot folder
- type: splot 'SAS.dat' using 3:4:5:1 with labels
- spin the plot around and find some informative angles
To get large png files of the angles you like, say 3000x2000 pixels, try...
- type: set term png size 3000, 2000
- type: set output "SAS.png"
- type: splot 'SAS.dat' using 3:4:5:1 with labels
Obviously, these days, there is no real genetic gap between Northern and Southern Europe. Populations across the middle of the continent merge into their Northern and Southern neighbors, bridging the two. So all we have are places where genetic distances are somewhat more pronounced than expected based on geographic proximity, such as the Alpine zone. However, was this the case before the "barbarian" migrations of the early middle ages, when Germanic and Slavic tribes moved south in great numbers? There's no way to tell, because it's not possible to get samples from that era. But I thought I'd see what happens in an MDS analysis if I took away all the groups supposedly affected by these population movements, and just ran the Northern Europeans against Southern Europeans and Middle Easterners. So I dropped the French, Swiss, Hungarians and Romanians, and got this...
That's quite a gap. However, it's possible to create such things on many plots with a bit of selective sampling. Nevertheless, I really like the analysis for a number of reasons, including the fact that it also shows a clear difference within Northern Europe in terms of southern influence. The cline that runs from Germany to Finland and the Baltic States gels extremely well with the lack of Neolithic influence in the Northeast of Europe, but fairly high Neolithic input in Western Europe. Check out this map from Haak et al. 2010, showing the similarity of modern mtDNA to that of ancient Neolithic LBK sites in Europe.
It's remarkable how MDS plots of modern populations can pick up such distant signals of ancient migrations. Obviously, I can't really prove that this is indeed what we've got here, but I can certainly see a correlation. I recommend that readers view this plot in an interactive window, so you can manipulate it with your mouse. It's easy...
- download Gnuplot
- unpack Gnuplot
- download the MDS data EU_ME.dat
- place EU_ME.dat into the "binary" Gnuplot folder
- type: splot 'EU_ME.dat' using 3:4:5:1 with labels
- spin the plot around and find some informative angles
- type: set term png size 2000, 1500
- type: set output "EU_ME.png"
- type: splot 'EU_ME.dat' using 3:4:5:1 with labels
And voila, that should give you 2000x1500 pixel png screen caps of the plot. I urge all of my project members to give it a try, because I'll be producing many more interactive plots like this. I have a feeling they'll go off like the proverbial frog in a sock.
Citation...
Haak W, Balanovsky O, Sanchez JJ, Koshel S, Zaporozhchenko V, et al. (2010) Ancient DNA from European Early Neolithic Farmers Reveals Their Near
Eastern Affinities. PLoS Biol 8(11): e1000536. doi:10.1371/journal.pbio.1000536
Before you do anything else, please install Gnuplot and download this datasheet. Drop the SEU.dat file into your working directory (c:\Documents and Settings\your own username), then open the cmd prompt and type "gnuplot", minus the quotation marks, followed by "splot 'SEU.dat' using 3:4:5:1 with labels", again minus the quotation marks. Alternatively, put the SEU.dat file into the "binary" Gnuplot folder, which you'll see after unzipping Gnuplot. If all goes well, you should see the image below in an interactive terminal, which you'll be able to spin around with your mouse. For more info on how to get this up and running see here.
For those of you who don't mind getting a bit techy, starting today I'll be posting datasheets that can be turned into interactive genetic maps with just a couple of simple commands. Here's the first one, an MDS analysis of the Baltic region, Russia and Scandinavia. This is what you'll see after following the instructions below, except that you'll be able to spin the plot in any direction with your mouse.
First of all, you'll have to install Gnuplot on your machine. You can download the latest version of this application here. Unfortunately, I can't help individual users with the installation, but it's pretty easy, and I'm sure you guys can help each other if problems arise. Now, once you have Gnuplot, just open your command prompt window and type gnuplot. Go to Start > Command Prompt > type Gnuplot. This is what you should see...
Then drop the "Baltic.dat" sheet into your working directory (which is usually C:\Documents and Settings\Username), and type "splot 'Baltic.dat' using 3:4:5:1 with labels" just like on the screen cap below, and hit enter. Alternatively, put the Baltic.dat file into the "binary" Gnuplot folder, which you'll see after unzipping Gnuplot. Once the interactive terminal pops up, you'll be able to edit the viewing options by pressing the buttons in the top left corner. For example, hit "Apply autoscale" to resize the terminal to fit your screen.
Baltic Finns show about 5-6% per cent more North and/or East Eurasian influence than Central Europeans. This affects their results on inter-continental analyses, like PCA-MDS, where they overlap with North Russians, with whom they're not particularly close genetically, apart from the fact that they share about the same level of the aforementioned eastern affinity. Does it also affect their results in intra-European tests? In other words, would Finns turn into Scandinavians or maybe Balts if their North/East Eurasian segments were removed? Let's check that.
Below are two MDS plots of five test Finns (FI1, FI2, FI5, FI6 and FI7). The first plot was produced using a random batch of 166K SNPs from the filtered 238K SNP set that I use in most analyses. The second one was done with 166K SNPs not contained within segments identified by me as North Eurasian, East Eurasian or Amerindian in the five test samples. Removing the putative eastern segments does have an affect on the results, with the Finns clearly shifting from the North Russian cluster towards the other Europeans with 5-8% less North/East Eurasian affinity (BY = Belarus, HU = Hungary, LT = Lithuania, NO = Norway, SE = Sweden). However, in all other aspects the two plots are very similar.
So it's safe to say the prep work was relatively successful. The targeted segments made up 3% to 5% of the total genomes in the test samples, and usually showed highest affinity to North Eurasian references like the Nganassan and Dolgan. In the near future, I plan to reconstruct some significant tracts made up of such eastern segments, sourced from various European and West Asian groups, to analyze more precisely their origins. If possible, it'd be really interesting to reconstruct entire genomes and run them on MDS plots.
Anyway, let's now have a look what sort of effect the two different sets of SNPs have on the behavior of the test samples on intra-European plots. The first pair of plots includes the test Finns, reference Finns with low level North/East Eurasian influence to begin with, as well as various groups that might have had an impact on the modern Finnish gene pool (Scandinavians, Balts, Russians and Germans).
- Random 166K SNPs
- Selected 166K SNPs
I can't see much of a change, except that the second plot looks more choppy, which is understandable, considering that the sampling of markers is less even. And now, here's another pair of plots, this time including a large set of Finns, from all over the country.
- Random 166K SNPs
- Selected 166K SNPs
Again, not a huge change. The test Finns don't shift their positions in the same meaningful way as on the inter-continental plots. Indeed, it's the same story on other plots, with different combinations of reference sets, which I won't bother posting here. It seems as if Finns are not significantly affected by their relatively increased North/East Eurasian affinity when placed on intra-European plots. Rather, their behavior on such plots, which often looks like fairly strong isolation compared to other Europeans, is affected by something else. That something else, I suspect, are fairly unique allele frequencies due to...isolation in the north followed by genetic drift.
The other interesting thing that came out of this exercise was a plot that, in my opinion, basically looks like a map of some of the major population movements and admixture events around the Baltic. Enjoy...
