PDA

View Full Version : The Problems with Kurd's Calculators



Xehanort
12-12-2017, 12:01 AM
Hi everyone,

Kurd's calculator results are very inconsistent. I hope he improves his algorithms and methods.

Here are my results for his K12, 23andMe V4:

STEPPE CULTURES 41.2%
KARASUK-E SCYTHIAN (2000-3000 years) 29.0%
ANDRONOVO-SRUBNAYA (3000-4000 years) 6.0%
YAMNAYA-AFANASIEVO-POLTAVKA (4000-5000 years) 6.2%

ANCIENT FARMERS 37.3%
WEST EUROPEAN FARMERS (4000-5000 years) 0.0%
LEVANT (4000-8000 years) 5.6%
NEOLITHIC-CHALCOLITHIC IRAN-CHG (5000-12000 years) 20.6%
EAST EUROPEAN FARMERS (5000-8000 years) 11.1%

SOUTHEAST EURASIAN 14.6%

EASTERN NON-AFRICANS 3.5%

AFRICAN 3.3%
EAST AFRICAN (modern) 3.3%

Here are my results for his K25:

C/SC/S ASIA 99.9%

Indian 51.3%

South Central Asian 33.1%
Central Asian 2 15.5%
Central Asian 1 0.0%

SIBERIAN & AMERINDIAN 0.1%
NATIVE AMERICAN 0.1%
Central Amerindian 0.0%
South Amerindian 0.0%
North Amerindian 0.0%

You would assume that based on my earlier results, I would be more Central Asian shifted on this calculator.

Now, K12 with AncestryDNA:

ANCIENT FARMERS 39.6%
WEST EUROPEAN FARMERS (4000-5000 years) 0.0%
LEVANT (4000-8000 years) 5.5%
NEOLITHIC-CHALCOLITHIC IRAN-CHG (5000-12000 years) 22.8%
EAST EUROPEAN FARMERS (5000-8000 years) 11.3%

STEPPE CULTURES 34.5%
KARASUK-E SCYTHIAN (2000-3000 years) 21.7%
ANDRONOVO-SRUBNAYA (3000-4000 years) 6.6%
YAMNAYA-AFANASIEVO-POLTAVKA (4000-5000 years) 6.2%

SOUTHEAST EURASIAN 17.5%

AFRICAN 3.9%
EAST AFRICAN (modern) 3.8%

EASTERN NON AFRICANS (modern) 3.7%

WESTERN EUROPEAN & SCANDINAVIAN HUNTER GATHERERS (4000-5000 years) 0.8%

And now, K25 with AncestryDNA:

C/SC/S ASIA 100.0%
Indian 52.9%
South Central Asian 28.1%
Central Asian 2 19.1%
Central Asian 1 0.0%

Based on these results, it is certain that Kurd's methods are not consistent. I would much rather go with David's (Eurogenes) methods and calculations. His are much more thorough and make much more sense to me.

David's fst4, for me:

Reference Populations:
Dai
Iran_N
Yamnaya_Samara

Results:
Dai: 0.182 (18.2%)
Iran_N: 0.455 (45.5%)
Yamnaya_Samara 0.362 (36.2%)

David's Admixture EBA_K7:

Reference Populations:
Steppe_EBA
Southeast_Asian
Anatolia_N
Sub-Saharan
Northeast_Asian
Euro_HG
Iran_N

Results:
Steppe_EBA: 0.239658 (24%)
Southeast_Asian: 0.099714 (10%)
Anatolia_N: 0.078884 (7.9%)
Sub-Saharan: 0.020627 (2.1%)
Northeast_Asian: 0.076556 (7.7%)
Euro_HG: 0.026345 (2.6%)
Iran_N: 0.458216 (45.8%)

From my analyses, it seems that David's calculators are much more consistent across the board in comparison to Kurd's. Especially considering that in Kurd's K12 calculators, Northern Europeans do not score as much Steppe as they should. Plus, the consistency between Kurd's calculators across different companies is practically non-existent. Perhaps, Kurd should make his methods more clear, because at the moment it seems that they are not as accurate as they should be. David also noted that the Steppe EBA K7 results are not as accurate as they should be for S/C Asians, because the calculator was designed for European populations, and hence, fst4 is much more accurate. For instance, Brahmins only score around 28% Steppe on average on his K7 calculator, but on his fst4 calculator, they score around 41%. At least David owns up to his errors, whereas Kurd (no offence, Kurd), claims that his calculators are the most accurate available on the market, when in reality, they are very inconsistent, especially when it comes to South Asians.

MonkeyDLuffy
12-16-2017, 06:37 PM
Kurd's calculator is optimized for 23andme V4. He has stated that multiple times that Ancestry DNA or FTDNA wont give the the right results with his calculators.

Xehanort
12-16-2017, 09:11 PM
Kurd's calculator is optimized for 23andme V4. He has stated that multiple times that Ancestry DNA or FTDNA wont give the the right results with his calculators.

Thanks for clarifying that, bro. But even with V4 I got the same results. It could just be my high Iran Neolithic component skewing my results. Though, we cannot be sure.

Sapporo
12-19-2017, 03:13 PM
Thanks for clarifying that, bro. But even with V4 I got the same results. It could just be my high Iran Neolithic component skewing my results. Though, we cannot be sure.

What do you mean by Central Asian shifted? South Central Asian toward Baloch/Brahui or Central Asian toward Tajikistani Tajiks? You're Pakistani Punjabi right? Are your results drastically different from other Pakistani Punjabis (Jatt, Arain, Rajput, etc.) on various Gedmatch calcs?

Xehanort
12-19-2017, 07:09 PM
What do you mean by Central Asian shifted? South Central Asian toward Baloch/Brahui or Central Asian toward Tajikistani Tajiks? You're Pakistani Punjabi right? Are your results drastically different from other Pakistani Punjabis (Jatt, Arain, Rajput, etc.) on various Gedmatch calcs?

Oh, Poi and MonkeyDLuffy Got results which were a bit shifted towards Central Asians and I am a Gujjar. Central Asian shifted towards Tajiks and Pamiris. Yeah. No, they're not, not in the least bit.

Kurd
12-19-2017, 07:12 PM
Hi everyone,

Kurd's calculator results are very inconsistent. I hope he improves his algorithms and methods.

Here are my results for his K12, 23andMe V4:

STEPPE CULTURES 41.2%
KARASUK-E SCYTHIAN (2000-3000 years) 29.0%
ANDRONOVO-SRUBNAYA (3000-4000 years) 6.0%
YAMNAYA-AFANASIEVO-POLTAVKA (4000-5000 years) 6.2%

ANCIENT FARMERS 37.3%
WEST EUROPEAN FARMERS (4000-5000 years) 0.0%
LEVANT (4000-8000 years) 5.6%
NEOLITHIC-CHALCOLITHIC IRAN-CHG (5000-12000 years) 20.6%
EAST EUROPEAN FARMERS (5000-8000 years) 11.1%

SOUTHEAST EURASIAN 14.6%

EASTERN NON-AFRICANS 3.5%

AFRICAN 3.3%
EAST AFRICAN (modern) 3.3%

Here are my results for his K25:

C/SC/S ASIA 99.9%

Indian 51.3%

South Central Asian 33.1%
Central Asian 2 15.5%
Central Asian 1 0.0%

SIBERIAN & AMERINDIAN 0.1%
NATIVE AMERICAN 0.1%
Central Amerindian 0.0%
South Amerindian 0.0%
North Amerindian 0.0%

You would assume that based on my earlier results, I would be more Central Asian shifted on this calculator.

Now, K12 with AncestryDNA:

ANCIENT FARMERS 39.6%
WEST EUROPEAN FARMERS (4000-5000 years) 0.0%
LEVANT (4000-8000 years) 5.5%
NEOLITHIC-CHALCOLITHIC IRAN-CHG (5000-12000 years) 22.8%
EAST EUROPEAN FARMERS (5000-8000 years) 11.3%

STEPPE CULTURES 34.5%
KARASUK-E SCYTHIAN (2000-3000 years) 21.7%
ANDRONOVO-SRUBNAYA (3000-4000 years) 6.6%
YAMNAYA-AFANASIEVO-POLTAVKA (4000-5000 years) 6.2%

SOUTHEAST EURASIAN 17.5%

AFRICAN 3.9%
EAST AFRICAN (modern) 3.8%

EASTERN NON AFRICANS (modern) 3.7%

WESTERN EUROPEAN & SCANDINAVIAN HUNTER GATHERERS (4000-5000 years) 0.8%

And now, K25 with AncestryDNA:

C/SC/S ASIA 100.0%
Indian 52.9%
South Central Asian 28.1%
Central Asian 2 19.1%
Central Asian 1 0.0%

Based on these results, it is certain that Kurd's methods are not consistent. I would much rather go with David's (Eurogenes) methods and calculations. His are much more thorough and make much more sense to me.

David's fst4, for me:

Reference Populations:
Dai
Iran_N
Yamnaya_Samara

Results:
Dai: 0.182 (18.2%)
Iran_N: 0.455 (45.5%)
Yamnaya_Samara 0.362 (36.2%)

David's Admixture EBA_K7:

Reference Populations:
Steppe_EBA
Southeast_Asian
Anatolia_N
Sub-Saharan
Northeast_Asian
Euro_HG
Iran_N

Results:
Steppe_EBA: 0.239658 (24%)
Southeast_Asian: 0.099714 (10%)
Anatolia_N: 0.078884 (7.9%)
Sub-Saharan: 0.020627 (2.1%)
Northeast_Asian: 0.076556 (7.7%)
Euro_HG: 0.026345 (2.6%)
Iran_N: 0.458216 (45.8%)

From my analyses, it seems that David's calculators are much more consistent across the board in comparison to Kurd's. Especially considering that in Kurd's K12 calculators, Northern Europeans do not score as much Steppe as they should. Plus, the consistency between Kurd's calculators across different companies is practically non-existent. Perhaps, Kurd should make his methods more clear, because at the moment it seems that they are not as accurate as they should be. David also noted that the Steppe EBA K7 results are not as accurate as they should be for S/C Asians, because the calculator was designed for European populations, and hence, fst4 is much more accurate. For instance, Brahmins only score around 28% Steppe on average on his K7 calculator, but on his fst4 calculator, they score around 41%. At least David owns up to his errors, whereas Kurd (no offence, Kurd), claims that his calculators are the most accurate available on the market, when in reality, they are very inconsistent, especially when it comes to South Asians.

If you had bothered to read my article at http://www.eurasiandna.com/2017/04/04/ancient-eurasia-admixture-calculator-update/ my methodology and the flaws with other calculators would have been clear to you, the only problem is to understand the rather technical aspects you would need to increase your knowledge on bioinformatics and genetics several fold because my articles and analysis and innovative methods require it.

Hobbyists admittedly will have a very hard time grasping the analysis and methodology I use, and I think everyone will agree that the analysis I do and articles I write are far more technical than anything put out by other calculator makers. I mean do you know anyone else who can genotype raw DNA sequences using various pipelines such as I have done http://www.eurasiandna.com/2017/10/02/diploid-genotyping-low-medium-coverage-ancient-dna/

If relatives are so genetically uniform, then why does IBS show us otherwise. Gedmatch calculators artificially minimize this variation. It is NOT real otherwise IBS would have supported it.

My advise to you is to carefully read those articles from my website and brush up on bioinformatics so that perhaps my algorithms will become more clear to you


Edit: If your problem is with the difference between your Ancestry and 23 data, I have numerous times stated 23 data will give the most accurate results, simply because there is much less overlapping markers between FTDNA and Ancestry and my tests. Do you think I was joking every time I posted that ?

Xehanort
12-19-2017, 09:01 PM
If you had bothered to read my article at http://www.eurasiandna.com/2017/04/04/ancient-eurasia-admixture-calculator-update/ my methodology and the flaws with other calculators would have been clear to you, the only problem is to understand the rather technical aspects you would need to increase your knowledge on bioinformatics and genetics several fold because my articles and analysis and innovative methods require it.

Hobbyists admittedly will have a very hard time grasping the analysis and methodology I use, and I think everyone will agree that the analysis I do and articles I write are far more technical than anything put out by other calculator makers. I mean do you know anyone else who can genotype raw DNA sequences using various pipelines such as I have done http://www.eurasiandna.com/2017/10/02/diploid-genotyping-low-medium-coverage-ancient-dna/

If relatives are so genetically uniform, then why does IBS show us otherwise. Gedmatch calculators artificially minimize this variation. It is NOT real otherwise IBS would have supported it.

My advise to you is to carefully read those articles from my website and brush up on bioinformatics so that perhaps my algorithms will become more clear to you


Edit: If your problem is with the difference between your Ancestry and 23 data, I have numerous times stated 23 data will give the most accurate results, simply because there is much less overlapping markers between FTDNA and Ancestry and my tests. Do you think I was joking every time I posted that ?

I used 23andMe data. Sorry I might have seemed to harsh here, though, I wanted to note that your other test had something completely different for me. I was listed as 41.2% Steppe, mostly Karasuk, and had very little Southeast Eurasian, so I assumed that I would get much more Central Asia like results here, but that is not the case. Could you explain to me why my results very significantly? I also did David's two recent tests, and my results were more or less in line with what was predicted with K12. Quite frankly, I am not at all angry or mad at you or anything. I just do not understand the variation here because Gujarati C and Gujarati D are heavily ASI, and yet, I don't show these ASI markers in your other tests. This is the problem I was having here, as to why my Indian component skyrocketed, and my Steppe component plummeted. David has a good method too, as good as yours I would say, there's just a lot of confusion here. David even plotted me correctly and I clustered with Sindhis, not Punjabis, surprisingly. Just wondering why there was a huge variation, and I never used FtDNA, not once lol. I used 23andMe and AncestryDNA.

Kurd
12-20-2017, 07:56 AM
I used 23andMe data. Sorry I might have seemed to harsh here, though, I wanted to note that your other test had something completely different for me. I was listed as 41.2% Steppe, mostly Karasuk, and had very little Southeast Eurasian, so I assumed that I would get much more Central Asia like results here, but that is not the case. Could you explain to me why my results very significantly? I also did David's two recent tests, and my results were more or less in line with what was predicted with K12. Quite frankly, I am not at all angry or mad at you or anything. I just do not understand the variation here because Gujarati C and Gujarati D are heavily ASI, and yet, I don't show these ASI markers in your other tests. This is the problem I was having here, as to why my Indian component skyrocketed, and my Steppe component plummeted. David has a good method too, as good as yours I would say, there's just a lot of confusion here. David even plotted me correctly and I clustered with Sindhis, not Punjabis, surprisingly. Just wondering why there was a huge variation, and I never used FtDNA, not once lol. I used 23andMe and AncestryDNA.



The K12 and K26 are totally different tests and I don't understand how you are correlating a certain % of Karasuk in the K12 to a certain % of C Asian in the K29. The BA Karasuk themselves are not exactly today's Turkmen or Tajikstanis. They best should be thought of something Okunevo type with a heavy dose of Siberian and some ancient Altaian, ie an eastern Steppe BA population. In the K12 they tend to proxy for ANE, Steppe, Siberian, and ancient Altaian, and you should not be correlating that with a certain % of Tajikstani or Turkmen in the K29 because that is not how it works. I don't understand how you have unilaterally decided how much ANE or Steppe or Siberian or ancient Altaian the various modern references of the K29 represent.

Individuals with the highest Karasuk scores in the K12 included SC and C Asians (and this includes SC Asians who score low Siberian and E Asian on modern calculators so I may have uncovered some connection between SC Asians and ancient E Steppe with this test) and S Asians who are SC/C Asian shifted such as Gujjars. I mean your combined score of 49% SC-C Asian in the K29 isn't exactly trivial, so your complaint about my calculators, which BTW may be considered innovative in certain aspects is unfounded ( there is a reason why I stopped making Gedmatch calculators a while back). Again, I strongly recommend that you read the articles on my website and brush up on fundamentals of genetics and bioinformatics.

Xehanort
12-20-2017, 09:03 AM
The K12 and K26 are totally different tests and I don't understand how you are correlating a certain % of Karasuk in the K12 to a certain % of C Asian in the K29. The BA Karasuk themselves are not exactly today's Turkmen or Tajikstanis. They best should be thought of something Okunevo type with a heavy dose of Siberian and some ancient Altaian, ie an eastern Steppe BA population. In the K12 they tend to proxy for ANE, Steppe, Siberian, and ancient Altaian, and you should not be correlating that with a certain % of Tajikstani or Turkmen in the K29 because that is not how it works. I don't understand how you have unilaterally decided how much ANE or Steppe or Siberian or ancient Altaian the various modern references of the K29 represent.

Individuals with the highest Karasuk scores in the K12 included SC and C Asians (and this includes SC Asians who score low Siberian and E Asian on modern calculators so I may have uncovered some connection between SC Asians and ancient E Steppe with this test) and S Asians who are SC/C Asian shifted such as Gujjars. I mean your combined score of 49% SC-C Asian in the K29 isn't exactly trivial, so your complaint about my calculators, which BTW may be considered innovative in certain aspects is unfounded ( there is a reason why I stopped making Gedmatch calculators a while back). Again, I strongly recommend that you read the articles on my website and brush up on fundamentals of genetics and bioinformatics.

Alright, thanks bro. Sorry for the speculations on my part. Best wishes! :) Thanks for explaining and clarifying things for me.

eolien
12-20-2017, 09:08 AM
...

Individuals with the highest Karasuk scores in the K12 included SC and C Asians (and this includes SC Asians who score low Siberian and E Asian on modern calculators so I may have uncovered some connection between SC Asians and ancient E Steppe with this test) and S Asians who are SC/C Asian shifted such as Gujjars. I mean your combined score of 49% SC-C Asian in the K29 isn't exactly trivial, so your complaint about my calculators, which BTW may be considered innovative in certain aspects is unfounded ( there is a reason why I stopped making Gedmatch calculators a while back). Again, I strongly recommend that you read the articles on my website and brush up on fundamentals of genetics and bioinformatics.


Dear Kurd,

I have 1+1 question about your calculator, perhaps you can try to reply when you have time:

1 - Does your calculator or any calculator take into account only the derived alleles or all the SNPs, since there are 700K data which is reduced to 100K, I guess there is a selection going on.

1b - what happens when 2 reference populations have the same % of the SNP allele, which one does the calculator assign (if for example albanian-greek and sicilian both have a derived allele for a particular SNP?

I have also 2 suggestions which might sound very ignorant or out of place, then discard them

2- instead of assigning an allele to a component, could we apply probability and give weights according to components: Let's say a SNP allele has 10% in one component, 50% in component 2, and 80% in component 3. I guess my allele is assigned to component 3 as you explained before. What if we give 0.8/0.1+0.5+0.8 to that component instead but also for example 0.1/0.1+0.5+0.8 to component 1 etc. along these lines, it is just a an idea at its initial stage.

2b - alternatively (or consequently) we give the compoenent results not as % but as a confidence range.

2c- what if we choose the SNPs for the calculator as equally distributed across the chromosomes because there might be some clusters. Otherwise I cannot explain my results derived from my parents.

thanks in advance

Sapporo
12-20-2017, 01:30 PM
I used 23andMe data. Sorry I might have seemed to harsh here, though, I wanted to note that your other test had something completely different for me. I was listed as 41.2% Steppe, mostly Karasuk, and had very little Southeast Eurasian, so I assumed that I would get much more Central Asia like results here, but that is not the case. Could you explain to me why my results very significantly? I also did David's two recent tests, and my results were more or less in line with what was predicted with K12. Quite frankly, I am not at all angry or mad at you or anything. I just do not understand the variation here because Gujarati C and Gujarati D are heavily ASI, and yet, I don't show these ASI markers in your other tests. This is the problem I was having here, as to why my Indian component skyrocketed, and my Steppe component plummeted. David has a good method too, as good as yours I would say, there's just a lot of confusion here. David even plotted me correctly and I clustered with Sindhis, not Punjabis, surprisingly. Just wondering why there was a huge variation, and I never used FtDNA, not once lol. I used 23andMe and AncestryDNA.

I think your SE Eurasian was underestimated on the previous calculator since you score rather uniformly with other Pakistani Punjabis on other Harappa/ANE K7/Gedmatch calculators from the results I've seen. You scored 14.6%, which is less than khana's Mohmand Pakhtun sample from FATA (16.80%) and close to Rukha (13.57%) who is half Afghan Pashtun/half Panjshiri Tajik (and has South Asian scores similar to Southern Pashtuns/Kandaharis). Now, if you scored lower or similar South Asian scores to the Mohmand individual and Rukha on a consistent basis across calculators, I wouldn't question its accuracy.

Your 14.6% is much lower than myself and other Punjabis (Pakistani and Indian) who scored mostly between 20-23%. Your higher ENA (3.5%) in comparison to myself (1.40%) may have compensated slightly but I'm not completely sure.

http://www.anthrogenica.com/showthread.php?12411-Pakistani-Gujjar-Results-Harappa

Now, with regards to the K25 and K29 calculators, the Indian component samples are Gujarati C and D. Both are heavily West Eurasian admixed populations in comparison to South Indian tribal groups like Paniyas. So, perhaps, your CHG and South Eurasian affinities like the component and hence an elevated score. Each individual has different affinities to modern and ancient populations. In MDL's case, he also scored noteworthy Eastern Euro and West Asian scores, which I completely lack here (other than 1.7% Baltic) and his Indian score may have elevated/adjusted (47% vs my 32%) to compensate.

If he had scored 7-8% Eastern Euro and 5% West Asian while scoring an Indian score very close to my own, it would indicate he is far more Northwestern/Central Asian shift/affinity than myself, which wouldn't be consistent with our results on other Gedmatch calculators or even Kurd's previous calculators.

Kurd
12-20-2017, 03:48 PM
Dear Kurd,

I have 1+1 question about your calculator, perhaps you can try to reply when you have time:

1 - Does your calculator or any calculator take into account only the derived alleles or all the SNPs, since there are 700K data which is reduced to 100K, I guess there is a selection going on.

1b - what happens when 2 reference populations have the same % of the SNP allele, which one does the calculator assign (if for example albanian-greek and sicilian both have a derived allele for a particular SNP?

I have also 2 suggestions which might sound very ignorant or out of place, then discard them

2- instead of assigning an allele to a component, could we apply probability and give weights according to components: Let's say a SNP allele has 10% in one component, 50% in component 2, and 80% in component 3. I guess my allele is assigned to component 3 as you explained before. What if we give 0.8/0.1+0.5+0.8 to that component instead but also for example 0.1/0.1+0.5+0.8 to component 1 etc. along these lines, it is just a an idea at its initial stage.

2b - alternatively (or consequently) we give the compoenent results not as % but as a confidence range.

2c- what if we choose the SNPs for the calculator as equally distributed across the chromosomes because there might be some clusters. Otherwise I cannot explain my results derived from my parents.

thanks in advance

Interesting questions.

1- Reference and test samples should be pruned to maximize overlapping SNPs for all samples for maximum accuracy. For example in the K29 the genotype rate is >99% for 23 V3 and V4 meaning that there is almost no gaps in overlap between the tester and the references meaning that there are no segments of the genome where the tester only overlaps some components and not others which would result in a bias (score lowering) on components for which there is less segment overlap (Ancestry, FTDNA, LivingDNA, 23 V5, etc). To get to a genotype rate of >99% translates into around 103K overlapping markers with my dataset.

1b- I think you mean same MAF (minor allele frequency-intrapopulation). This does in fact happen at some positions especially with components such as Scandanvian-NW European which are separated by low genetic distances. No one knows the answer except maybe Alexander (person who put out the ADMIXTURE program). I am assuming those positions are not used in the calculation.

I also believe and this is purely based on my very extensive understanding of the program due to 100s of runs and my extensive knowledge of bioinformatics (you will not see this anywhere online) that the tester is sometimes the tie-breaker for some positions where MAF is the same between 2 closely related pops. To test for this I did a test using a parent-child trio to determine what % of positions in the child are homozygous reference/major or alternate/minor allele (0/0 or 1/1) while both parents are 0/1. Using a common sub-set of markers it turned out to be 10%.
So I believe it is for those positions that you occasionally see a child score a component outside the range of the parents. For example, the parents score 0 and 10% W Asian, whereas the child scores 0% W Asian, but instead scores 5% Caucasian (where both parents are 0% Caucasian).
In other word, the fact that the child is 1/1 in a position where the parents are 0/1 is where the child's extra allele acts as a tie-breaker for 2 pops with a similar MAF at that position.

2c- I checked your results with your parents and they are fine. You fall within the range of your parents for most components. The components where you score <1.5% where your parents are 0 is likely due to noise or could less likely be due to the above.

Xehanort
12-24-2017, 09:49 AM
I think your SE Eurasian was underestimated on the previous calculator since you score rather uniformly with other Pakistani Punjabis on other Harappa/ANE K7/Gedmatch calculators from the results I've seen. You scored 14.6%, which is less than khana's Mohmand Pakhtun sample from FATA (16.80%) and close to Rukha (13.57%) who is half Afghan Pashtun/half Panjshiri Tajik (and has South Asian scores similar to Southern Pashtuns/Kandaharis). Now, if you scored lower or similar South Asian scores to the Mohmand individual and Rukha on a consistent basis across calculators, I wouldn't question its accuracy.

Your 14.6% is much lower than myself and other Punjabis (Pakistani and Indian) who scored mostly between 20-23%. Your higher ENA (3.5%) in comparison to myself (1.40%) may have compensated slightly but I'm not completely sure.

http://www.anthrogenica.com/showthread.php?12411-Pakistani-Gujjar-Results-Harappa

Now, with regards to the K25 and K29 calculators, the Indian component samples are Gujarati C and D. Both are heavily West Eurasian admixed populations in comparison to South Indian tribal groups like Paniyas. So, perhaps, your CHG and South Eurasian affinities like the component and hence an elevated score. Each individual has different affinities to modern and ancient populations. In MDL's case, he also scored noteworthy Eastern Euro and West Asian scores, which I completely lack here (other than 1.7% Baltic) and his Indian score may have elevated/adjusted (47% vs my 32%) to compensate.

If he had scored 7-8% Eastern Euro and 5% West Asian while scoring an Indian score very close to my own, it would indicate he is far more Northwestern/Central Asian shift/affinity than myself, which wouldn't be consistent with our results on other Gedmatch calculators or even Kurd's previous calculators.

I don't get much ASI in any calculators though. In Harappa I got 31.17, and in David's calculator I got 18. I think it's just the Neolithic Iran affinities.

Sapporo
12-24-2017, 01:52 PM
I don't get much ASI in any calculators though. In Harappa I got 31.17, and in David's calculator I got 18. I think it's just the Neolithic Iran affinities.
I'm not suggesting you do.

However, if you're referring to Harappa, you mean South Indian, which is a "mixed" modern component (based on various modern South Indian samples used by Zack) from before the discovery of ancient genomes. It's not an accurate gauge of actual ASI ancestry. Anyways, Rukha and the Mohmand Pakhtun get significantly less SI than you on Harappa. As for David's calculator, I assume you're referring to ANE K7. It was one of the more popular Gedmatch calculators after the discovery of the Siberian Mal'ta boy and ANE (Ancestral North Eurasian) ancestry. The ASE or Ancestral South Eurasian on there is still not a proper proxy for ASI but certainly better than Harappa. Moreover, both Rukha and the Mohmand individual score less ASE than you there as well. My point is that your SE Eurasian score on Kurd's K25 calculator isn't consistent with your "South Eurasian" scores across various Gedmatch calculators when compared to others but to be fair, they're testing different things as well.

I'll also note that I score slightly less SI than you on Harappa and less ASE on ANE K7 as well yet I scored significantly more SE Eurasian than you on that particular Geneplaza calculator. Every calculator will show something different but I'd imagine we would should be scoring rather similar percentages most of the time.

Rinema
12-30-2017, 06:37 PM
Sorry, my message is just an essay (i'm new)

MonkeyDLuffy
12-30-2017, 09:34 PM
Sorry, my message is just an essay (i'm new)

Where's the essay?

poi
12-31-2017, 02:02 AM
I'm not suggesting you do.

However, if you're referring to Harappa, you mean South Indian, which is a "mixed" modern component (based on various modern South Indian samples used by Zack) from before the discovery of ancient genomes. It's not an accurate gauge of actual ASI ancestry. Anyways, Rukha and the Mohmand Pakhtun get significantly less SI than you on Harappa. As for David's calculator, I assume you're referring to ANE K7. It was one of the more popular Gedmatch calculators after the discovery of the Siberian Mal'ta boy and ANE (Ancestral North Eurasian) ancestry. The ASE or Ancestral South Eurasian on there is still not a proper proxy for ASI but certainly better than Harappa. Moreover, both Rukha and the Mohmand individual score less ASE than you there as well. My point is that your SE Eurasian score on Kurd's K25 calculator isn't consistent with your "South Eurasian" scores across various Gedmatch calculators when compared to others but to be fair, they're testing different things as well.

I'll also note that I score slightly less SI than you on Harappa and less ASE on ANE K7 as well yet I scored significantly more SE Eurasian than you on that particular Geneplaza calculator. Every calculator will show something different but I'd imagine we would should be scoring rather similar percentages most of the time.

I think xehanort was referring to Davidski's new paid calculator. The components were Dai, Yamnaya_Samara, and Iran_N.

Rinema
01-01-2018, 02:57 PM
MonkeyDLuffy I have difficulties to send my messages :(

I have to do it several times and this is what I get :
"The following errors occurred with your submission
An error has occurred Rinema! You are not allowed to post links."

Camulogène Rix
01-01-2018, 03:06 PM
Where's the essay?

She means an attempt

Xehanort
01-05-2018, 09:26 AM
I'm not suggesting you do.

However, if you're referring to Harappa, you mean South Indian, which is a "mixed" modern component (based on various modern South Indian samples used by Zack) from before the discovery of ancient genomes. It's not an accurate gauge of actual ASI ancestry. Anyways, Rukha and the Mohmand Pakhtun get significantly less SI than you on Harappa. As for David's calculator, I assume you're referring to ANE K7. It was one of the more popular Gedmatch calculators after the discovery of the Siberian Mal'ta boy and ANE (Ancestral North Eurasian) ancestry. The ASE or Ancestral South Eurasian on there is still not a proper proxy for ASI but certainly better than Harappa. Moreover, both Rukha and the Mohmand individual score less ASE than you there as well. My point is that your SE Eurasian score on Kurd's K25 calculator isn't consistent with your "South Eurasian" scores across various Gedmatch calculators when compared to others but to be fair, they're testing different things as well.

I'll also note that I score slightly less SI than you on Harappa and less ASE on ANE K7 as well yet I scored significantly more SE Eurasian than you on that particular Geneplaza calculator. Every calculator will show something different but I'd imagine we would should be scoring rather similar percentages most of the time.

Thanks, but on ANE K7 I score around 16% ASE, I was referring to David's latest calculator, fst 4, not his other one. Appreciate the clarifications.

Xehanort
01-05-2018, 09:28 AM
MonkeyDLuffy I have difficulties to send my messages :(

I have to do it several times and this is what I get :
"The following errors occurred with your submission
An error has occurred Rinema! You are not allowed to post links."

Are you half Kabyle? Your SSA percentages are very low for a North African, or even half North African.

Sapporo
01-05-2018, 09:59 AM
Thanks, but on ANE K7 I score around 16% ASE, I was referring to David's latest calculator, fst 4, not his other one. Appreciate the clarifications.

Okay, I'm not familiar with this calculator but you would need to compare your results to other Punjabis/NW South Asians as well and look for consistency across calculators.

Xehanort
01-05-2018, 04:48 PM
Okay, I'm not familiar with this calculator but you would need to compare your results to other Punjabis/NW South Asians as well and look for consistency across calculators.

Alright, thanks bro. Appreciate it. Any news about Rakhigarhi?