Fennoscandian BGA MDS plot update 59 participants

New update of the MDS plot with 59 participants. The total number of genomewide SNP have been reduced to 289k SNP as we now include 2 FamilyFinder participant.

The structure remains much the same as earlier

North-East European, not Siberian (updated)

In earlier posts I have indicated that dimension 1 in the later MDS plots was of Sibieran origin. This I assumed based on similarities with the clustering with or without Siberians. I have now done a new review with the Siberian and conclude thats not the case. This is the plot in question I have published earlier. In the plot below dimension 1 (horisontal) I suggested to be a Siberian (left) vs European (right) dimension while dimension 2 I suggested to be an western European (bottom) and eastern European dimension (top). After new insights this may need some more modifications:

I did a new run with the current panel. I added 4 unrelated Sibierans (Ngaans and Dolgan) into the panel. This gave the following MDS plot:

Impossible to see the details here but the general impression is that Siberians is a fairly distant population. Dimension 1 here is the horisontal axis where the extreme left is Siberian while the extreme right is European. Dimension 2 appears to be an internal European dimension with the North Italians at one extreme while the The Finns appear on the other extreme suggesting the dimension is a north-south dimension. The Saamis appears to stick out somewhat from the Finns. In this image its hard to see the details so lets us zoom in.

As we can see here in detail the Italians have the lowest Siberian influence. It appears to be in a continuity all the way downward and increase more to left as we go down the vertical axis. The Scandinavians is inside partly in the Belorussian and Hungarian cluster. However from this part the Siberian influence separate into the Vologda Russians and the Lithuanians. Most Finns (with some internal variations) appears to have similar Siberian influence as the Vologda Russians.  The Saami appears to continue further left at the same downward level as part of the Finns.

However as the plot shows above the European dimension (D2) isnt more informative than showing an North-South influence. We have an additional dimension 3 and when we compare it do dimension 2 we get a very familiar MDS plot to the one seen at the top just rotated left 90 degrees:

So basically what the earlier plot showed was that dimension was a north-south axis or more preciously a south-west vs north-east genetic and geographic distribution while dimension 2 again separate it into an west and east genetic and geographic distribution or to be more precise a north-west vs south-east genetic and geographic distribution. Below dimension 1 represented geographically.

Dimension SNP (and variation):

D1: 51 706 SNP (10,6%)
D2: 17 341 SNP (3,6%)
D3:  9 758 SNP (2,0%)

Geographic MDS maps of Fennoscandia and Europe (updated)

There exists some similarity between Finns and Lithuanians not seen in Saamis, so when adding the Lithuanians the Finns moved out of the North-Saami cluster compared to the previous analysis. Therefore Lithuanians where added. Also FI5 and FI6 where 2nd cousins pulled down the Finns into the North-Saami cluster. Removing each indepentently resulted in the same effect suggesting the relatedness had an effect in the analysis. FI6 was therefore removed.

The MDS plots can be confusing for some. I therefore made a map showing the geographic average distributions for the 3 main components and it may show an intriging history. Dimension 1 accounts for the most of the variation, while dimension 2 account for the second largest genetic variantion while dimension 3 account for the third largest genetic variation. Therefore the dimensions do not acount equally when represented in a 3D view.

The map appears to show Saamis as extremes in all three dimensions.

This is how much each dimension explains the genetic variation in each dimension:

 D1: 20 604 SNP 4.2%
 D2: 10 417 SNP 2.1%
 D3: 8 560 SNP 1.8%
Dimension Total: 8.1%

D1: Saamis, Finns and Vologda Russians appears to make a far eastern bridge. Earlier analysis seem to suggest that this is a Siberian component seen in Europe. It appears absolutely lowest among the Italians. Note that the Italians appears to be "isolated" in Europe in this context with much higher rates seen in close populations. Scandinavians appear closest to Belorussians and Hungarians in this dimension.

D2 also appears very high among the Saami and seem to have a western European distribution as it has high frequency also among Finns, Swedes and Norwegians. It is also higher in western Europe as seen among the French and Italians. However it drops very dramatically from Finns to Lithuanians, Estonia and Vologda Russian. The level increases again in more western east-European populations like Hungarians and Romanians.

D3 also appears very high among the Saami but shows an intriguing pattern. Its very low among Scandinavians, Lithuanians and Estonians, however the next highest occurrences are found among the Italians, then the Romanians and the Vologda Russians and then finally Finns just somewhat above intermediate levels like the French, Hungarians and Belorussians. 

May it be the "eastern component" researcher Tambets mentions in her abstract for the coming paper about genetic of the uralic speakers? If yes why is it also found second highest among Italians?

Also the local opposite clustering of this component only seen among Norwegians, Swedes, Lithuanians and Estonians are intriguing suggesting something in common.

As the result indicate the Saami outlierness appears to be the sum of these 3 different dimensions or genetic components. Italians appears to be opposite outlier in D1, Lithuanians appears to be opposite outlier in D2 and Swedes appears to be opposite outlier in D3.

Fennoscandia BGA regional MDS update (2nd update)

We got 5 new project members since last update. SWE17, SWE18, SWE19. FI12 and FI13. The interpretations are much the same as given in earlier posts. We can see that SWE14 and SWE17 are outside the main Scandinavian cluster pulling toward the eastern European populations. SWE19 appears to cluster well into the Scandinavain cluster. Interesting that FI12 a southern Finn appears close to the Estonian ES1 halfway towards the Vologda Russians and the Belorussians. FI13 have ancestry from Savo and cluster in the Karelian cluster.  

We also see some slight movements of others after adding more individuals. This is normal as adding more individuals influence the positioning of all other individuals especially the closer ones to new added individuals.

Fennoscandian Mutation Sharing Matrix

All participants have received the Mutation Sharing Matrix of currently known unique haplotype mutations map for Fennoscandia shared with two or more individuals. These mutations is currently only known in Fennoscandia.

Its highly recommended to use Excel 2007 or later when using. Enclosed Excel 2007 and Excel 2003 versions.

X-axis - Individuals
Y-axis - Clusters

How to use:

1. Find your code at the top.
2. Hit the filter and choose "1".
3. You now have a overview over at what cluster and to what other individuals you share mutations with.
4. To the far right is a indicative distribution of the cluster etnical distribution. Please use with extreme caution in regards to interpretation.

Other things:

* Individuals of known partly Saami ancestry is indicated with blue colour.
* Individuals of known partly Saami ancestry is labeled as Norwegians and Swedes.
* Interpretation of mutation patterns must be done with extreme caution. There are many possible problems like etnical categorisations for haplotypes that is also shared outside ethical boundaries but within geographical boundaries. The mutation could be "native" but also arrived trough later migration.
* Widespread distributions may indicate higher age for the mutation cluster or it may indicate a common immigrant who hit the genetic jackpot.

If you have not received the sheets please contact me on email.

Anders

Accurate method for pinpointing autosomal ancestry?

In the earlier post I reported to have found 2 haplotype clusters seen in 6 individuals each. I here look closer at 1 of those 2 clusters to see what can be inferred from them.

The core haplotype

The haplotype identified occured in 6 individuals: 3 Finns, 2 Swedes and 1 Norwegian (no Saamis) and no one else in the population panel. The core of the haplotype is GC with the markers rs6683734 and rs4649296. These are together mutations that can only be found among Fennoscandians. However the spread is very limited only occuring in 6 of 88 haplotypes (or "3" of 44 individuals).

The extented haplotype

The haplotype can be furter extended for all with these markers: rs16858853 rs12087818 rs701177 rs16858884 rs701176 rs701173 rs11800619 rs10489806 rs7546115 rs10910120 rs1033322 rs12124323 rs17752790 rs6697791 rs12140862 rs12130755 rs6683734 rs4649296. The physical distance between the first and last is 75kb or 0.075 Mb. The individual haplotype appears to have gone trough recombination if extending the haplotype further.

The extended haplotypes founder haplotype appears to have accumulated several mutations that appears to have some local variation suggest it has a age and history in Fennoscandia seen in total. Two Finns FI11 and FI12 appears to share what appears to be the founder haplotype. FI10 have accumulated two mutations not seen in anywhere else. The SWE1 and SWE13 haplotypes appears to have separate history from the founderhaplotype seen in the two Finns. Here the SWE1 appears to be the founder while the SWE13 appears as a subfounder. The one Norwegian haplotype NO5 appears to have a history closest to the Finn haplotype.

The mutation diversity suggest a history even its widespread. The founder haplotype is seen in Finland while 4 of 6 mutations is seen in Swedes and a Norwegian. The Swedes even have a sub-node haplotype. This suggest that Sweden is the oldest location for this haplotype with Finland as runner up. It may be youngest in Norway but belong to the Scandinavian cluster in other analysis so it may be grouped with Swedes. More samples could possibly enlight the question of origin.

The ancestral haplotype

It would be of great interest to identify the true ancestral haplotype to the above described haplotype in Continental-Europe to be able to infer the "point" of entry to Fennoscandia.

I managed to extract this possible ancestral haplotypes. Two from Spain and 1 from the French Basque. As seen from these it appears like third last and last SNP mutated in Fennoscandia from AGA to GGC. The Norwegian's AGC may be due to back mutations, error or it could be transition between the third last and and last SNP. If correct this could mean the point of entry was Norway but it is inconsistent with extended haplotype having the lowest diversity but not if we group Swedes and Norwegians.

The SNP before the last tree is the common founder haplotype. I attempted to find sign of diversity for the Iberians for this part but didnt manage to detect any at first screening. The zero diversity of the Iberian group for this part compared to the Fennoscandian suggest the Fennoscandian haplotype are older. The difference is mutated positions for the third last and last SNP.

SUMMARY:  As seen from this posting trying to reconstruct a autosomal haplotype could be complex but give accurate and informative genetic history. The core haplotype appears to be simple and effective in proving Fennoscandinavian ancestry despite its tiny size, while the extended haplotype can provide clues to its age and history.

This method using core haplotypes with limited distributed mutations appears to be very accurate in pinpointing ancestry

Autosomal Haplotype Clustering Patterns - Actual or Error? (updated)

I have found this pattern of haplotype clustering within individuals from Norway, Sweden and Finland on Chr 1 using 38.5k SNP:


Unique haplotypes not shared with others - 1 344 (typical between 15 to 50 per individual)
Haplotypes shared between 2 ind - 157
Haplotypes shared between 3 ind - 30
Haplotypes shared between 4 ind - 3
Haplotypes shared between 5 ind - 2
Haplotypes shared between 6 ind - 2


As this shows widespread haplotype clusters are much rarer than those shared with only two individuals, but the "unique" haplotype clusters appears to be absolutely highest at the individual level.

This raises the questions why its like this. I suspect its the following reasons:

1. The effect of recombination splitting or killing haplotypes. However the maximum haplotype size in clusters is 500 SNP. Redusing it to 100 SNP only reduced to 1324 unique haplotype clusters. Reducing to max 10 SNP only reduced to 1189 unique haplotype clusters. Reducing to 5 SNP reduced only to 962 unique haplotype clusters. If reducing to 2 SNP only 277 unique haplotype clusters.
2. The effect of limited population data. Its possible more individuals and populations would reduce the number of unique haplotype clusters.
3. The effect of undetected errors in the genotypes. However no correlation between high unique haplotypes found in individuals and high detected genotype error rate for these.
4. The effect of incorrect phasing as the result of errors in genotype or/and ordinary phasing error as result of the model used.
5. The effect of haplotype or mutation extinction. Recent individual haplotypes or mutations have limited spread generally, while older haplotype clusters or mutations have larger geographic spread.

So what I infer from this is that these unique haplotype clusters is rather small and not very large. These numbers have been generated from software made for finding genetic diseases from haplotypes where you mark individuals with certain traits cases and check them vs the controls. If there is any haplotype strongly associated with a trait the associated haplotype is found. These haplotypes are usually not very large. Just check the SNP used by 23andme health section.. So is also the cases with these haplotypes.

The software do for many haplotypes infer parent-child relationships between them indicating that haplotype mutations are in the picture at least when I check at the individual level.