PDA

View Full Version : My results: What to make of an unusual component?



therrien.joel
09-01-2017, 04:28 AM
Hello,

Uploaded my 23andMe raw data to DNALand along with GEDmatch with interesting results. Interesting in that DNALand came way closer to getting my ancestry correct at least according to what I have been told (1/2 French Canadian, 1/4 English, 1/8 Irish and 1/8 Finnish).

DNALand result: 100% West Eurasian
65% Northwest European
8% Southwestern European
6% Sardinian
11% Finnish
6.9% Ashkenazi
2.9% Central Asian (1.5% Kalash and the rest "ambiguous")

Heck of a lot better than what 23andMe came up with

Reading up on this forum and trying the Eurogenes Jtest, I can be fairly confident the Ashkanazi is a fluke. I even don't put much weight on the specific assignment of Kalash. But the almost 3% Central Asian has me curious. I can't rule it out definitively but am cautious about reading tea leaves as well. One funny thing is that using the Eurogenes k12b calculator I get similarly "reasonable" results:

Population
Western European 40.05
Siberian -
East African -
West Central Asian 3.07
South Asian -
West African -
Caucasus 5.96
Finnish 12.16
Mediterranean 15.29
Southwest Asian 0.26
North European 22.50
East Asian 0.70

Conveniently this calculator has no oracle nor a modern population spreadsheet, so I can only take it at face value, which again is not all that far off.

But there it is again: West Central Asian 3%. Other calculators pick up west asian, sometimes Indian (not Harappaworld though 0% S. Indian and 10.5% Baloch) but I account for that by assuming most europeans see some of those.

I'll say this, just to get it off my chest because I would be lying if I claimed it wasn't partly why this is getting me curious... All my life I have had subtly odd comments made to me. My high school nickname was Hadji (from Johnny Quest and no, I tan well but am white. They said it was my face). Through college I had a few people ask me if I was part asian. Even my wife, who is mainland Chinese, made a comment once in a while. All that really doesn't amount to a hill of beans as far as proof that this is or isn't just statistical noise.

I'd love to hear some comments on what anyone thinks or any suggestions on exploring this further, if it is even worth exploring.

therrien.joel
09-03-2017, 02:39 AM
As a sanity check, I went back over the other calculators on GEDmatch that had population spreadsheets so I could work out the population numbers that I should have based on my expected ancestry (Oracle is useless b/c of the 1/8'th components). One or two check out, but the vast majority indicate that the calculator is yielding population percentages that do not match what is expected. And the mismatches are all occurring for populations from the same region (Central Asia).

This is happening for MDLP, Eurogenes, Dodecad, Gedroisia and actually Harrapa World too. Perhaps DNA Land is actually catching something?

wombatofthenorth
09-03-2017, 03:16 AM
My mom and I don't get WCA on E K12b but my dad gets like a percent and a half. He may have a trace of Balkan, maybe some sort of Slovenian, greater Romanian/western Ukrainian. So maybe it is a sign of a bit of Balkan/Ukrainian/even Russian or something like that?
Maybe hints of Mongol invasions or something??
He also has a trace, maybe 2% Finnish.

wombatofthenorth
09-03-2017, 03:17 AM
Finns get ancient Northeast Asian signals or some other signals from the East. Sometimes they have an Asian phenotype clearly show through.

therrien.joel
09-03-2017, 05:31 PM
That's a good point. I have vague recollections of my grandmother mentioning some Balkan heredity, but that is unsubstantiated. I tried adding in a small bit of populations from that area. It gets closer but still doesn't balance very well, some excess goes away but some other populations stay high. I need to play around with it some more. Still seems to point more east. If it is Balkans maybe Romani would explain that. Otherwise I wish I had a getter idea if any of my British ancestors were in India.

Amerijoe
09-03-2017, 07:03 PM
Joel, welcome. Almost two years ago, when I first started testing, my ydna was place under the heading of Z2123, simply mystified me. A native of Paisley, Scotland being classified as Asian. Here I am, pale white, blue eyes, red hair being Asian, must be a mistake. Upon further research, I discover there is more than one side of the coin. From an autosomal perspective, I'm very Scot/Irish, high 90% across all commercial testing platforms. Needless to say with the assistance of many knowledgeable experts and much help from Anthrogenica members, I was able to verify the Asian relationship.

It was was suggested that the best approach to solve this mystery was through a BigY test. Last Dec. that test was completed, resulting in a cache of Asian matches. Farthest matches India, Sri Lanka, Saudi Arabia, Palestine, Iran, Iraq, but rather list them all, here is a map of the match area covering most of my ydna matches. http://www.anthrogenica.com/attachment.php?attachmentid=18538&stc=1 Most of my ancient autosomal genes have been replaced producing a typical Scot/Irish appearance. Here is my K12b results.
http://www.anthrogenica.com/attachment.php?attachmentid=18539&stc=1

JohnHowellsTyrfro
09-03-2017, 07:30 PM
This sort of question comes up pretty often. It's very probably "ancient" and the "Asian" Kalash etc, is the closest modern population owing to relative genetic isolation. Sardinia, Basque etc. can fall into that category too. I've had Ashkenazi on some tests ( I'm British) and it's probably confused with something else in terms of early origins. Thread below. John

http://www.anthrogenica.com/showthread.php?11851-Basque-Sardinian-SW-Asian-(Burusho-etc-)-Welsh-or-Irish

Amerijoe
09-03-2017, 07:49 PM
FTDNA has me at about 2% Ashkenazi. There are subtle DNA differences between my YDNA Jewish matches and my Arab matches which adds to the analysis confusion. I call these minority ancestries, tag-alongs, many picked up on the way to becoming you. Where, how and when are what population geneticists are also asking. With the combination of more universal and ancient testing, a clearer picture will hopefully materialize.

therrien.joel
09-04-2017, 01:27 AM
This sort of question comes up pretty often. It's very probably "ancient" and the "Asian" Kalash etc, is the closest modern population owing to relative genetic isolation. Sardinia, Basque etc. can fall into that category too. I've had Ashkenazi on some tests ( I'm British) and it's probably confused with something else in terms of early origins. Thread below. John



That's very informative, thank you. I agree, the most probably answer is that it is "ancient". But to not leave a stone unturned, since I don't have the benefit of a well documented ancestry, it is interesting to see how close to genealogically relevant time this might be.

One thing to bear in mind is that I am not relying on the raw population percentages that a calculator yields since as the thread that you linked to and many other sources make it clear that unusual trace populations can crop up for a variety of reasons (ancient connections for example). What I am doing instead is basically manually calculating the percentages of each population I would expect to see for my heritage. If I come up short, that's not a big issue since genetic transmission is not perfectly 50/50 from parents to child. But if I get an excess, then I pay some attention. In many cases I am finding double the expected amount. I have also looked at the chromosome paintings and do tend to find runs of between 3-5 cM where the only population assigned is Central Asian. Far from a smoking gun, but enough to keep me digging.

I have also done some 1-1 comparisons with GEDmatch and have found some matches with kit numbers of individuals that claim to be Romani, Iranian and Armenian (loosening the genealogically rigorous criteria to 3cM and 500SNP). Similar criteria with people from populations that aren't from that region do not match at all.

Certainly I realize all this will ultimately yield no more than an interesting story to tell over drinks, but that doesn't make it any less interesting!

Mixed
09-04-2017, 01:30 AM
23&me did not process quite a bit of my dna.

therrien.joel
09-04-2017, 01:32 AM
Thank you Amerijoe. Very cool! I wish I was in a position to know that digging into things like the y-DNA would yield something interesting.

therrien.joel
09-04-2017, 01:35 AM
23&me did not process quite a bit of my dna.

I am honestly starting to wonder how well most of the commercial testing companies handle people who are mixed. Since matching to someone that is 100% X-nationality is not the same as someone with a pinch of this and a dash of that.

JohnHowellsTyrfro
09-04-2017, 06:56 AM
That's very informative, thank you. I agree, the most probably answer is that it is "ancient". But to not leave a stone unturned, since I don't have the benefit of a well documented ancestry, it is interesting to see how close to genealogically relevant time this might be.

One thing to bear in mind is that I am not relying on the raw population percentages that a calculator yields since as the thread that you linked to and many other sources make it clear that unusual trace populations can crop up for a variety of reasons (ancient connections for example). What I am doing instead is basically manually calculating the percentages of each population I would expect to see for my heritage. If I come up short, that's not a big issue since genetic transmission is not perfectly 50/50 from parents to child. But if I get an excess, then I pay some attention. In many cases I am finding double the expected amount. I have also looked at the chromosome paintings and do tend to find runs of between 3-5 cM where the only population assigned is Central Asian. Far from a smoking gun, but enough to keep me digging.

I have also done some 1-1 comparisons with GEDmatch and have found some matches with kit numbers of individuals that claim to be Romani, Iranian and Armenian (loosening the genealogically rigorous criteria to 3cM and 500SNP). Similar criteria with people from populations that aren't from that region do not match at all.

Certainly I realize all this will ultimately yield no more than an interesting story to tell over drinks, but that doesn't make it any less interesting!

Yes you are right. I suppose I'm slightly different in that my ancestry is British so I know pretty much what it is. I've gone down the route of exploring Hun, Finnish or Roma ancestry to explain "unexpected" small percentages, now I'm pretty sure in my own case, it's just "old". :) Good luck in tracking it down. John

therrien.joel
09-05-2017, 02:01 AM
Yes you are right. I suppose I'm slightly different in that my ancestry is British so I know pretty much what it is. I've gone down the route of exploring Hun, Finnish or Roma ancestry to explain "unexpected" small percentages, now I'm pretty sure in my own case, it's just "old". :) Good luck in tracking it down. John

Thanks!

I'm inclined to agree with that assessment. I took the liberty of checking your own GEDmatch kit number using the chromosome painter. The Asian markers that you have are pretty fragmented, which is in agreement with that ancestry being old. Hope you don't mind, I was curious to see how that looked for someone who was more certain of their heritage than myself and the difference stood out; I have much larger segments that are assigned to only Central Asian populations. Juuust outside the theoretical "reach" of a genealogical match which makes it fairly tantalizing.

spruithean
09-05-2017, 02:56 AM
Thanks!

I'm inclined to agree with that assessment. I took the liberty of checking your own GEDmatch kit number using the chromosome painter. The Asian markers that you have are pretty fragmented, which is in agreement with that ancestry being old. Hope you don't mind, I was curious to see how that looked for someone who was more certain of their heritage than myself and the difference stood out; I have much larger segments that are assigned to only Central Asian populations. Juuust outside the theoretical "reach" of a genealogical match which makes it fairly tantalizing.

I'm trying to figure out what constitutes fragmented in this sense. On my chromosome paintings on various calculators my South Asian result seems far larger than fragmented. I'm not sure why. I'm also not sure if it is recent or if it is ancient. I will have to find a way to compare.

L1983
09-05-2017, 05:43 AM
Have you tested a parent? Only ask because I tend to get this component, think I got Kalash and indo-iranian in dna land. In the k12b I get 5.16, but mum gets just over 4 and dad gets just over 2. Never looked at matches, but would assume lowering to 3cM would be very inanncurate and produce false matches. I've always just assumed it's a blip in my dna, but not knowing all of my ancestry it does make you wonder. You're welcome to look at my gedmatch. Just pm me.

JohnHowellsTyrfro
09-05-2017, 06:08 AM
Thanks!

I'm inclined to agree with that assessment. I took the liberty of checking your own GEDmatch kit number using the chromosome painter. The Asian markers that you have are pretty fragmented, which is in agreement with that ancestry being old. Hope you don't mind, I was curious to see how that looked for someone who was more certain of their heritage than myself and the difference stood out; I have much larger segments that are assigned to only Central Asian populations. Juuust outside the theoretical "reach" of a genealogical match which makes it fairly tantalizing.

No problem at all, I appreciate your thoughts. I'm not great at understanding the technicalities, so it just helps provide me with some answers and I've been trying to answer that question for quite some time and glad that it helped you a little too. :) John

JohnHowellsTyrfro
09-05-2017, 06:16 AM
I'm trying to figure out what constitutes fragmented in this sense. On my chromosome paintings on various calculators my South Asian result seems far larger than fragmented. I'm not sure why. I'm also not sure if it is recent or if it is ancient. I will have to find a way to compare.

My ancestry is pretty much European British going way back, mostly Welsh and some English as far as I know and it seems my Asian is ancient. If you want to compare with me feel free. T034724 John

JohnHowellsTyrfro
09-05-2017, 06:20 AM
Have you tested a parent? Only ask because I tend to get this component, think I got Kalash and indo-iranian in dna land. In the k12b I get 5.16, but mum gets just over 4 and dad gets just over 2. Never looked at matches, but would assume lowering to 3cM would be very inanncurate and produce false matches. I've always just assumed it's a blip in my dna, but not knowing all of my ancestry it does make you wonder. You're welcome to look at my gedmatch. Just pm me.

No parents or siblings alive. The reason I suspect it isn't just noise is because it's consistent on different tests/calculators (sometimes with different "labels") and it's not just me. Of course there could be some other explanation like coincidence or something, but it's possible there may be more to it I think. John

L1983
09-05-2017, 06:37 AM
No parents or siblings alive. The reason I suspect it isn't just noise is because it's consistent on different tests/calculators (sometimes with different "labels") and it's not just me. Of course there could be some other explanation like coincidence or something, but it's possible there may be more to it I think. John

Sorry, that was to OP. I forgot to do reply with quote. Not woken up properly! I can only go by my experience, which of course is just my experience, but when I went by just my results I was convinced there was some unknown history lurking somewhere that I didn't know about. Even the creator of the MDLP calculator thought I had a great grandparent from Iran or Afghanistan when he looked at my raw data. Convinced my parents to get tested and... well, nothing above the noise categories anyhow.

JohnHowellsTyrfro
09-05-2017, 06:59 AM
Sorry, that was to OP. I forgot to do reply with quote. Not woken up properly! I can only go by my experience, which of course is just my experience, but when I went by just my results I was convinced there was some unknown history lurking somewhere that I didn't know about. Even the creator of the MDLP calculator thought I had a great grandparent from Iran or Afghanistan when he looked at my raw data. Convinced my parents to get tested and... well, nothing above the noise categories anyhow.

My very non-technical thoughts (a little knowledge is a dangerous thing) is that could this relate more to the background genetic "soup" which could have certain regional characteristics? I understand what we inherit from recent ancestors is very much a lottery, However within a regional population perhaps the probabilities of picking up something "unusual" might be higher? I'm thinking here in terms of regional populations rather than recent ancestry, bearing in mind how many ancestors we all have as we go back through the generations? Just thoughts. :) John

therrien.joel
09-05-2017, 03:08 PM
Have you tested a parent? Only ask because I tend to get this component, think I got Kalash and indo-iranian in dna land. In the k12b I get 5.16, but mum gets just over 4 and dad gets just over 2. Never looked at matches, but would assume lowering to 3cM would be very inanncurate and produce false matches. I've always just assumed it's a blip in my dna, but not knowing all of my ancestry it does make you wonder. You're welcome to look at my gedmatch. Just pm me.

Very good idea. Not sure if I would be able to convince either of them to do it, but perhaps I'll get the curious at some point.

I agree, the 3cM is a low threshold, so one has to be careful of false positives. That was also why I did not lower the default SNP threshold because that really makes almost everyone seem related. Actually the hits I got were in the range of 4 even one 5 cM match, which at least is a bit better. But again, I am not putting a lot of weight on this approach. maybe only if I had a statistically significant batch of kits to compare I might be able to say something.

Given that a 1% difference is within the likely statistical variance, the fact that you and you mom are close to that doesn't exactly disprove a link.

therrien.joel
09-05-2017, 03:29 PM
I'm trying to figure out what constitutes fragmented in this sense. On my chromosome paintings on various calculators my South Asian result seems far larger than fragmented. I'm not sure why. I'm also not sure if it is recent or if it is ancient. I will have to find a way to compare.

Wish I could post pictures it would be easier to explain, but have not crossed whatever threshold is required for that yet.

Basically, if you go to the chromosome painting option and look at the different populations' colors you should see regions that are almost totally one color. Those are the regions that are more reliably linked to that one population because the others rarely carry that allele. If you look at the length of those segments in millions of base-pairs you can ballpark the number of centiMorgans (~1M BP = 1cM with an understanding that this varies within even a single chromosome, so it is only a rough gauge). If the number is above 7, then you have a decent chance it is real. Below that you are getting more iffy, but if you see a solid single color it's at least a better bet that it is derived from that population but too far back to find a paper trail. If almost no area shows a single color or it is very small then you can say it is much more fragmented. How big the the largest segment that is painted South Asian? I don't get huge hits on South Asian but on West Asian, Caucus and West Central Asian I can find segments that range from 4-7 MBP. Just close enough to make me think it isn't a fluke.

L1983
09-05-2017, 03:30 PM
Very good idea. Not sure if I would be able to convince either of them to do it, but perhaps I'll get the curious at some point.

I agree, the 3cM is a low threshold, so one has to be careful of false positives. That was also why I did not lower the default SNP threshold because that really makes almost everyone seem related. Actually the hits I got were in the range of 4 even one 5 cM match, which at least is a bit better. But again, I am not putting a lot of weight on this approach. maybe only if I had a statistically significant batch of kits to compare I might be able to say something.

Given that a 1% difference is within the likely statistical variance, the fact that you and you mom are close to that doesn't exactly disprove a link.

I remember I got a match to an Iranian guy at 5cM which was basically dismissed by another group outright as I had lowered the threshold. I don't have the patience to do more digging with matches, unfortunately and I lack the knowledge to know what's significant in chromosome view. What's your gedmatch out of interest. I'd like to see how we compare if you don't mind?

therrien.joel
09-05-2017, 03:40 PM
My very non-technical thoughts (a little knowledge is a dangerous thing) is that could this relate more to the background genetic "soup" which could have certain regional characteristics? I understand what we inherit from recent ancestors is very much a lottery, However within a regional population perhaps the probabilities of picking up something "unusual" might be higher? I'm thinking here in terms of regional populations rather than recent ancestry, bearing in mind how many ancestors we all have as we go back through the generations? Just thoughts. :) John

I think as we progress in developing these analysis kits we are eventually going to have to come to terms with increasing the amount of unknowable information we have to accept. I know this flies in the face of us all wanting and believing that eventually we will be able to nail down every last SNP and have a high resolution map of our ancestry. But in reality, you are correct, it is a soup and we can't really make better progress until we accept that a significant chunk of our genome just won't yield the information we seek. Until we accept that and get more honest about the limits of these calculators, we have to accept the high noise levels that they have. Outright dismissing an unusual result is as bad as accepting it at face value. I think for those who want to really know, we have to invest the time and do the manual searches.

As a scientist, this is proving to be a very nice test of my training. Here I have an ambiguous result in which I have some interest and more likely than not some bias. Can I apply the analytical skills I have developed over the years in a responsible fashion to get the most truthful answer I can? Keeping in mind that knowing techniques very well means you also can subconsciously exploit their weaknesses to bias the results.

JohnHowellsTyrfro
09-05-2017, 05:54 PM
I think as we progress in developing these analysis kits we are eventually going to have to come to terms with increasing the amount of unknowable information we have to accept. I know this flies in the face of us all wanting and believing that eventually we will be able to nail down every last SNP and have a high resolution map of our ancestry. But in reality, you are correct, it is a soup and we can't really make better progress until we accept that a significant chunk of our genome just won't yield the information we seek. Until we accept that and get more honest about the limits of these calculators, we have to accept the high noise levels that they have. Outright dismissing an unusual result is as bad as accepting it at face value. I think for those who want to really know, we have to invest the time and do the manual searches.

As a scientist, this is proving to be a very nice test of my training. Here I have an ambiguous result in which I have some interest and more likely than not some bias. Can I apply the analytical skills I have developed over the years in a responsible fashion to get the most truthful answer I can? Keeping in mind that knowing techniques very well means you also can subconsciously exploit their weaknesses to bias the results.

Yes you are right, I'm just speculating of course because it's interesting. :) If eventually more people test, we might be able to draw better conclusions about population groups and trends. The individual's result may reveal less maybe, it may not be indicating what you think it is but it could be indicating something. John

A Norfolk L-M20
09-05-2017, 05:57 PM
Joel, welcome. Almost two years ago, when I first started testing, my ydna was place under the heading of Z2123, simply mystified me. A native of Paisley, Scotland being classified as Asian. Here I am, pale white, blue eyes, red hair being Asian, must be a mistake. Upon further research, I discover there is more than one side of the coin. From an autosomal perspective, I'm very Scot/Irish, high 90% across all commercial testing platforms. Needless to say with the assistance of many knowledgeable experts and much help from Anthrogenica members, I was able to verify the Asian relationship.

It was was suggested that the best approach to solve this mystery was through a BigY test. Last Dec. that test was completed, resulting in a cache of Asian matches. Farthest matches India, Sri Lanka, Saudi Arabia, Palestine, Iran, Iraq, but rather list them all, here is a map of the match area covering most of my ydna matches. http://www.anthrogenica.com/attachment.php?attachmentid=18538&stc=1 Most of my ancient autosomal genes have been replaced producing a typical Scot/Irish appearance. Here is my K12b results.
http://www.anthrogenica.com/attachment.php?attachmentid=18539&stc=1

And my L-SK1414 being another example. An English person with a haplogroup more likely to be found between the Levant and Afghanistan. But all autosomal evidence washed away.

therrien.joel
09-05-2017, 06:19 PM
I remember I got a match to an Iranian guy at 5cM which was basically dismissed by another group outright as I had lowered the threshold. I don't have the patience to do more digging with matches, unfortunately and I lack the knowledge to know what's significant in chromosome view. What's your gedmatch out of interest. I'd like to see how we compare if you don't mind?

I would say that if you lower the threshold, then it is better to have multiple matches and importantly 100% non-matches in populations that you aren't indicated to belong to. Of course it is also an issue of how to interpret the results; with some holding a hard line on what is genealogically relevant, i.e. what could be found by a paper trail. I respect that attitude because as John righty said, a little information can be dangerous.

Since I am not yet above the threshold to PM you and it doesn't matter too much anyway, here is my GEDmatch kit number: M108553 I'd be interested to see what you get.

therrien.joel
09-05-2017, 06:21 PM
And my L-SK1414 being another example. An English person with a haplogroup more likely to be found between the Levant and Afghanistan. But all autosomal evidence washed away.

Absolutely! That was basically good luck that an interesting tidbit like that managed to take shelter in the y-chromosome.

geebee
09-05-2017, 07:22 PM
I am honestly starting to wonder how well most of the commercial testing companies handle people who are mixed. Since matching to someone that is 100% X-nationality is not the same as someone with a pinch of this and a dash of that.

I actually think 23andMe didn't do too badly with my analysis. At the broadest level, they say I'm 97.9% "European", 2.0% "East Asian and Native American" (which is essentially all Native American), and 0.1% "unassigned". That pretty much fits with my family tree, which includes somewhat distant Native American ancestry through both my maternal grandparents.

Breaking the "European" down a bit more precisely, it's 90.7% "Northwestern European" in the form of 43.5% "British & Irish", 22.4% "French & German", 0.4% "Finnish", and 24.4% "broadly Northwestern European". I also have 5.1% "Southern European" as 1.9% Iberian, 0.8% "Italian", and 2.3% "broadly Southern European"; and 2.2% "broadly European".

Probably the biggest puzzle for me here is why the Southern European is actually as low as it is. It's traceable to my most recent immigrant ancestors, who are three of my maternal grandmother's grandparents. Both of her grandfathers immigrated from the Spanish island of Minorca -- separately, as children with their families; and one of her grandmothers immigrated a child from Alsace-Lorraine. But that means my grandmother should be half Spanish, my mother should be around a fourth, and I should be an eighth. That's more like 12.5%, not 5.1% -- but obviously I either didn't inherit as much DNA from my Minorca 2nd great grandfathers as I might have, or else they would not themselves have tested as 100% "Southern European".

The "East Asian & Native American" is broken down as 2.0% "Native American" and <0.1% "broadly East Asian & Native American", and my five siblings all have similar numbers.

One complaint I have is that even though these "broadly" categories are listed with their own percentages as subcategories -- just like, say, "British & Irish", clicking on them doesn't result in highlighting just those segments. Instead, if you click on "broadly Northwestern European" you get the same highlighting as for the entire "Northwestern European" category. But this makes no sense, because for me "Northwestern European" is 90.7% but "broadly Northwestern European" is only 24.4%.

Providing separate highlighting would allow me to see those segments which could not be identified more finely than just as "broadly Northwestern European", and it might allow me to figure out what more specific assignment could make the most sense.

I'll use one of my sister's results to try to clarify what I mean. Her "East Asian & Native American" is a bit higher than mine at 2.5%, but it also includes a tiny amount that is identified as specifically "East Asian" -- in fact, as "Mongolian". But Mongolian ancestry makes no sense for my sister, and because it has its own highlighting I can easily see that it's shown as just a tiny segment in the middle of a much larger "Native American" segment. In all likelihood, it is just Native American, though it looks a bit different to 23andMe's algorithm.

By the same token, if my "broadly European" segments only appeared within or near "Southern European" segments, that would tell me that's probably what they actually are. But if I click on the category, everything European will be highlighted at once.

Sorry for wandering a bit off-topic.

I suppose the bottom line is really that you have the best chance of figuring out an "odd" component if it's shown in segment form, rather than just as percentages; and if you're able to place it within a larger context. After all, it isn't really percentages we inherit from our ancestors, it's actual DNA segments.

therrien.joel
09-06-2017, 01:36 AM
I suppose the bottom line is really that you have the best chance of figuring out an "odd" component if it's shown in segment form, rather than just as percentages; and if you're able to place it within a larger context. After all, it isn't really percentages we inherit from our ancestors, it's actual DNA segments.

100% agree with you on that!

JohnHowellsTyrfro
09-06-2017, 06:43 AM
I actually think 23andMe didn't do too badly with my analysis. At the broadest level, they say I'm 97.9% "European", 2.0% "East Asian and Native American" (which is essentially all Native American), and 0.1% "unassigned". That pretty much fits with my family tree, which includes somewhat distant Native American ancestry through both my maternal grandparents.

Breaking the "European" down a bit more precisely, it's 90.7% "Northwestern European" in the form of 43.5% "British & Irish", 22.4% "French & German", 0.4% "Finnish", and 24.4% "broadly Northwestern European". I also have 5.1% "Southern European" as 1.9% Iberian, 0.8% "Italian", and 2.3% "broadly Southern European"; and 2.2% "broadly European".

Probably the biggest puzzle for me here is why the Southern European is actually as low as it is. It's traceable to my most recent immigrant ancestors, who are three of my maternal grandmother's grandparents. Both of her grandfathers immigrated from the Spanish island of Minorca -- separately, as children with their families; and one of her grandmothers immigrated a child from Alsace-Lorraine. But that means my grandmother should be half Spanish, my mother should be around a fourth, and I should be an eighth. That's more like 12.5%, not 5.1% -- but obviously I either didn't inherit as much DNA from my Minorca 2nd great grandfathers as I might have, or else they would not themselves have tested as 100% "Southern European".

The "East Asian & Native American" is broken down as 2.0% "Native American" and <0.1% "broadly East Asian & Native American", and my five siblings all have similar numbers.

One complaint I have is that even though these "broadly" categories are listed with their own percentages as subcategories -- just like, say, "British & Irish", clicking on them doesn't result in highlighting just those segments. Instead, if you click on "broadly Northwestern European" you get the same highlighting as for the entire "Northwestern European" category. But this makes no sense, because for me "Northwestern European" is 90.7% but "broadly Northwestern European" is only 24.4%.

Providing separate highlighting would allow me to see those segments which could not be identified more finely than just as "broadly Northwestern European", and it might allow me to figure out what more specific assignment could make the most sense.

I'll use one of my sister's results to try to clarify what I mean. Her "East Asian & Native American" is a bit higher than mine at 2.5%, but it also includes a tiny amount that is identified as specifically "East Asian" -- in fact, as "Mongolian". But Mongolian ancestry makes no sense for my sister, and because it has its own highlighting I can easily see that it's shown as just a tiny segment in the middle of a much larger "Native American" segment. In all likelihood, it is just Native American, though it looks a bit different to 23andMe's algorithm.

By the same token, if my "broadly European" segments only appeared within or near "Southern European" segments, that would tell me that's probably what they actually are. But if I click on the category, everything European will be highlighted at once.

Sorry for wandering a bit off-topic.

I suppose the bottom line is really that you have the best chance of figuring out an "odd" component if it's shown in segment form, rather than just as percentages; and if you're able to place it within a larger context. After all, it isn't really percentages we inherit from our ancestors, it's actual DNA segments.

I think we all tend to forget these are all just labels related to similarities with modern populations. If these small percentages relate to very old ancestry the modern label has less meaning to some extent. I'm British and get around 1% Native American/Amerindian/Arctic Amerindian on various tests. If it means anything at all in my own case, I guess it could relate to very early Asian origins or it could just be some sort of genetic coincidence but other "Asian" indicators in my results may mean it's part of that "package". :) John

therrien.joel
09-06-2017, 03:30 PM
I think we all tend to forget these are all just labels related to similarities with modern populations. If these small percentages relate to very old ancestry the modern label has less meaning to some extent. I'm British and get around 1% Native American/Amerindian/Arctic Amerindian on various tests. If it means anything at all in my own case, I guess it could relate to very early Asian origins or it could just be some sort of genetic coincidence but other "Asian" indicators in my results may mean it's part of that "package". :) John

Yes, that's an annoying "feature". I also suffer from the fact that my legend extends twice the length of the actual mapped chromosomes, so I can't even browse half of them. Not to mention 23andMe doesn't give you a very high resolution view.

therrien.joel
09-07-2017, 01:05 AM
So I think at this point I am inclined to believe that DNAland is on to something in regards to the Central Asian component they arrived at.

I spent the last few days finally making a "smart oracle" that could attempt to tease out which ethic groups best accounted for the discrepancy that I have consistently seen in my GEDmatch calculator results. Very briefly, what it does is start with your "known" admixture and then use a technique that we use in quantum physics called perturbation theory. Essentially it replaces a small percentage of your known heritage with each ethic group that a particular calculator has data on. It iterates through all possible combinations, identifying those that bring the percentages closest to what the calculator gave you. This was necessary for me because my known admixture is not compatible with the oracles.

Now the interesting results...
- Results from two calculators gave nonsense results (MDLP K16 and puntDNAL k13 global... though in fairness the latter's output might make some sense broadly speaking)
- Dodecad, Gedrosia and Harrapaworld all yielded more or less the same result: a list of ethnicities that fill in the discrepancies rather well. I put them on a map to make it more easy to see.

18587

They all are around 3% and interestingly, they all subtract from the French component. Funny enough, I did not know this until I looked it up tonight, but France did actually have a colony in India.

L1983
09-07-2017, 01:30 AM
So I think at this point I am inclined to believe that DNAland is on to something in regards to the Central Asian component they arrived at.

I spent the last few days finally making a "smart oracle" that could attempt to tease out which ethic groups best accounted for the discrepancy that I have consistently seen in my GEDmatch calculator results. Very briefly, what it does is start with your "known" admixture and then use a technique that we use in quantum physics called perturbation theory. Essentially it replaces a small percentage of your known heritage with each ethic group that a particular calculator has data on. It iterates through all possible combinations, identifying those that bring the percentages closest to what the calculator gave you. This was necessary for me because my known admixture is not compatible with the oracles.

Now the interesting results...
- Results from two calculators gave nonsense results (MDLP K16 and puntDNAL k13 global... though in fairness the latter's output might make some sense broadly speaking)
- Dodecad, Gedrosia and Harrapaworld all yielded more or less the same result: a list of ethnicities that fill in the discrepancies rather well. I put them on a map to make it more easy to see.

18587

They all are around 3% and interestingly, they all subtract from the French component. Funny enough, I did not know this until I looked it up tonight, but France did actually have a colony in India.

This is very cool and rather intriguing!

therrien.joel
09-23-2017, 02:47 AM
In case anyone was curious, I have also submitted my data to Gencove. They appear to use a similar if not the same algorithm as DNAland (Joe Pickrell is the CEO and he was involved in DNAland's development as fas I understand). A key difference is they use different reference populations, including more South Asians allowing them to give more details. Here is what they get for my results:

18941

18940

I also found by pure accident someone who originates from the Punjab area. This person and another person who is acknowledged as a 'likely fourth cousin from the same village' triangulate with me over three (short) segments on different chromosomes. So, sounds like this is probably real.