PDA

View Full Version : Online Pigmentation Predictor



Kurd
11-27-2017, 12:13 AM
I am in the process of putting together a pigmentation calculator that is a little more accurate than what is out there, with at least 5 pigment levels instead of the 3 level predictors you may have seen out there.

Predicting pigmentation is a little more difficult than eye color or hair color, as human pigmentation is controlled by many genes which act in different ways at different stages of melanin production. thus the accuracy of pigmentation predictors does not appear to be as high as eye or hair color prediction.

There does appear to be a free online pigmentation predictor that is a little more robust than what I have seen out there. It bases its output on 10 SNPs, 7 of which are available if your data is 23andMe V4. It is available at http://mathgene.usc.es/snipper/skinclassifier.html

Your participation will help me design something a little more accurate with a couple of more levels of pigmentation output.

There are 3 levels of each of the categories; white, intermediate, and black. You also get a prediction admixture result such as 98% White, 2% Intermediate, 0% Black. The prediction is for your untanned skin.

All you have to do is input your genotype at the 10 SNPs specified.

Please take a screen shot as I have done below, because simply posting your prediction admixture is not that informative to me.

Also try to rate from 1 to 5 the accuracy of the prediction, with 1 being very accurate.

To get the ball rolling here are the screen shots of the prediction for 2 northern Iraqi Kurds with whom I am personally familiar with and myself.


N Iraqi Kurd. If I had to pick a shade of white for this individual the right most sample (darkest) shade of White would be closest to their untanned skin.

https://i.imgur.com/x8YXzAC.jpg

Kurd
11-27-2017, 12:19 AM
Here is another northern Iraqi Kurd who i personally familiar with. The prediction for him is 94.36% white, 5.64% intermediate, which is a little bit lighter than the previous individual who was 89.33% white, 10.65% intermediate. I would say this indeed appears to be correct since this individual is a shade lighter than the previous one, and has lighter eyes (grreenish-hazel) and dark brown hair, whereas the previous one has brown eyes and black hair.


https://i.imgur.com/RaAF9dR.jpg

Kurd
11-27-2017, 12:26 AM
BTW, although ^^ is a shade lighter than the Kurd in post 1, I would still say that the right most shade of white is closest to his untanned skin.

This one is for myself. I would say I am closest to the left most shade (lightest) of intermediate (untanned)

https://i.imgur.com/TWZ21gH.jpg

surbakhunWeesste
11-27-2017, 12:43 AM
I hope I did this correct.
19991
19992
19993

I am reddish and not pale lol.

Anyway these predictions never work in my case. It was fun.

Kurd
11-27-2017, 12:58 AM
I hope I did this correct.
http://www.anthrogenica.com/attachment.php?attachmentid=19991&stc=1
http://www.anthrogenica.com/attachment.php?attachmentid=19992&stc=1
http://www.anthrogenica.com/attachment.php?attachmentid=19993&stc=1

I am reddish and not pale lol.

Anyway these predictions never work in my case. It was fun.

I agree their skin tone palate is pretty limited. Also, not sure how much the 3 missing SNPs from 23andme affect the result. Personally, I don’t think 10 genes is sufficient, I believe pigmentation is affected by at least 30 genes. I will try to incorporate at least hopefully 30 into my calculator . I’m sure there will be some for which this one is off. It’s not easy, but It’ll be interesting to see how various ethnicities do.

Do you mind also posting what the admixture prediction was

surbakhunWeesste
11-27-2017, 01:07 AM
I agree their skin tone palate is pretty limited. Also, not sure how much the 3 missing SNPs from 23andme affect the result. Personally, I donít think 10 genes is sufficient, I believe pigmentation is affected by at least 30 genes. I will try to incorporate at least hopefully 30 into my calculator . Iím sure there will be some for which this one is off. Itís not easy, but Itíll be interesting to see how various ethnicities do.

Do you mind also posting what the admixture prediction was

You mean this, my eyes hurt lol
19994

firemonkey
11-27-2017, 01:25 AM
If 23andMe only has 7 of the markers where do I get the rest?

Kurd
11-27-2017, 01:39 AM
If 23andMe only has 7 of the markers where do I get the rest?

You can’t, but it will work with the 7 markers. That’s what was used for the previous users. Just type nn for the ones you are missing

Kurd
11-27-2017, 01:45 AM
I believe untanned olive skin falls under white (darkest shade) because olive (untanned) is how I would describe the Kurd in post 1

firemonkey
11-27-2017, 01:53 AM
Hope this helps.

19995

surbakhunWeesste
11-27-2017, 01:54 AM
my dad's
you can cross check if I did it correctly with his raw data

19996
19997
19998

He likes to sun-tan but is pale naturally, does not really burn. I burn but I am reddish not pale.
really confusing

I hope my contributions help, given I did it correctly.

Kurd
11-27-2017, 02:02 AM
my dad's
you can cross check if I did it correctly with his raw data

http://www.anthrogenica.com/attachment.php?attachmentid=19996&stc=1
http://www.anthrogenica.com/attachment.php?attachmentid=19997&stc=1
http://www.anthrogenica.com/attachment.php?attachmentid=19998&stc=1

It seems very likely that the 7 SNPs are insufficient for accurate prediction especially for Asians. Although I haven’t seen your dad, this seems off especially when I compare with Firemonkey assuming he is average for British. It is possible that the study was based on Europeans mostly and thus more accurate for Europeans.

Kurd
11-27-2017, 02:05 AM
Hope this helps.

http://www.anthrogenica.com/attachment.php?attachmentid=19995&stc=1

Do you consider yourself average complexion for British, and if you had to pick one of the white shades which one would it be for your untanned skin?

surbakhunWeesste
11-27-2017, 02:09 AM
It seems very likely that the 7 SNPs are insufficient for accurate prediction especially for Asians. Although I haven’t seen your dad, this seems off especially when I compare with Firemonkey assuming he is average for British. It is possible that the study was based on Europeans mostly and thus more accurate for Europeans.

IKR, (checked firemonkey's result),my dad has fair 'ME' skin tone. I don't know how else to explain his natural complexion. We are special people, need special snps analysis lulz. Maybe more data contribution on this thread could shed some light.

Kurd
11-27-2017, 02:28 AM
IKR, (checked firemonkey's result),my dad has fair 'ME' skin tone. I don't know how else to explain his natural complexion. We are special people, need special snps analysis lulz. Maybe more data contribution on this thread could shed some light.

Agreed. The same applies to the 2 Iraqi Kurds I posted. Even the fairer one is a few shades darker than “average” British, with the one in post 1 of similar complexion and facial appearance as most Pashtuns and northern Pakistanis, yet they both have a higher “White” score than Firemonkey, again assuming that Firemonkey is not darker than S Italians

Cohort should have many Asians and include at least 30 genes.

firemonkey
11-27-2017, 12:06 PM
Do you consider yourself average complexion for British, and if you had to pick one of the white shades which one would it be for your untanned skin?

Using the Fitzpatrick scale I would say -2. White,fair. Using the Von Luschan scale- 7-13 light or light skinned European.

Sangarius
11-27-2017, 12:56 PM
https://abload.de/img/2017_11_27_13_26_06_rp1r5k.png

firemonkey
11-27-2017, 02:07 PM
I am surprised a Turk's predicted white admixture would be considerably more than mine.

Kurd
11-27-2017, 02:47 PM
I am surprised a Turk's predicted white admixture would be considerably more than mine.

Yes, and 2 Kurds and 2 Afghans. The results are likely more comparable only within geographical regions, W/S Asians with W/S Asians, Europeans with Europeans. I am working on something using a different principal and many more genes that will yield more accurate results than their predictor. We shall See....

firemonkey
11-27-2017, 03:18 PM
Have any other Europeans contributed to this thread to do a comparison?

anglesqueville
11-27-2017, 03:48 PM
Did it for my mom. She is very pale, in fact nearly depigmented (several operations for potentially cancerous tumors):
20006

firemonkey
11-27-2017, 03:58 PM
Did it for my mom. She is very pale, in fact nearly depigmented (several operations for potentially cancerous tumors):
20006

Wow! Much more white than me.

anglesqueville
11-27-2017, 04:06 PM
Wow! Much more white than me.

Perhaps indeed a little too white. :\

Dorkymon
11-27-2017, 04:16 PM
Mine, I don't get rosy unless I'm doing a physically demanding task or cold, but otherwise I'm quite pale.
I'd say that it was accurate for me.

https://i.imgur.com/P2QGGYY.png

firemonkey
11-27-2017, 04:31 PM
So far it seems my profile is the least white of the Europeans. Very much explained by the comparatively increased intermediate.

vettor
11-27-2017, 04:39 PM
I have 23andme not found # 5, 9 and 10 ..........the others are

GGGGCTAA..GGAAGG....

how do I write these non found in the above to predict?

Kurd
11-27-2017, 04:43 PM
I have 23andme not found # 5, 9 and 10 ..........the others are

GGGGCTAA..GGAAGG....

how do I write these non found in the above to predict?


Use nn for the 3 missing ones

vettor
11-27-2017, 04:48 PM
Use nn for the 3 missing ones

Did not work , regards

I have v3 version


rs1408799 12672097 C or T C / T

rs3829241 68855363 A or G G / G

mine above for the errors shown from program below

Error: SNP-Indel rs1408799 (number 3) has possible alleles AG but your profile has first allele C in this SNP-Indel (global position 5).
Error: SNP-Indel rs1408799 (number 3) has possible alleles AG but your profile has second allele T in this SNP-Indel (global position 6).
Error: SNP-Indel rs3829241 (number 8) has possible alleles CT but your profile has first allele G in this SNP-Indel (global position 15).
Error: SNP-Indel rs3829241 (number 8) has possible alleles CT but your profile has second allele G in this SNP-Indel (global positi

it also presented me with


SNP-Indel names rs10777129 rs13289 rs1408799 rs1426654 rs1448484 rs16891982 rs2402130 rs3829241 rs6058017 rs6119471
Alleles AG CG AG AG AG CG AG CT AG CG

Please curate your profile and/or training set.

vettor
11-27-2017, 05:09 PM
Did not work , regards

I have v3 version


rs1408799 12672097 C or T C / T

rs3829241 68855363 A or G G / G

mine above for the errors shown from program below

Error: SNP-Indel rs1408799 (number 3) has possible alleles AG but your profile has first allele C in this SNP-Indel (global position 5).
Error: SNP-Indel rs1408799 (number 3) has possible alleles AG but your profile has second allele T in this SNP-Indel (global position 6).
Error: SNP-Indel rs3829241 (number 8) has possible alleles CT but your profile has first allele G in this SNP-Indel (global position 15).
Error: SNP-Indel rs3829241 (number 8) has possible alleles CT but your profile has second allele G in this SNP-Indel (global positi

it also presented me with


SNP-Indel names rs10777129 rs13289 rs1408799 rs1426654 rs1448484 rs16891982 rs2402130 rs3829241 rs6058017 rs6119471
Alleles AG CG AG AG AG CG AG CT AG CG

Please curate your profile and/or training set.

I replaced my error ones with their options and got


This profile is 46.23 times more likely White than Intermediate, and 27,558 times more likely White than Black.

Predicted admixture: 97.88 % for White; 2.12 % for Intermediate; 0.00 % for Black.
Therefore, the profile should be White.

Kurd
11-27-2017, 05:38 PM
I replaced my error ones with their options and got


This profile is 46.23 times more likely White than Intermediate, and 27,558 times more likely White than Black.

Predicted admixture: 97.88 % for White; 2.12 % for Intermediate; 0.00 % for Black.
Therefore, the profile should be White.

so you did change CT to AG at rs1408799, and GG toCC at position 8, right?

Can you doa screen shot?

vettor
11-27-2017, 05:51 PM
so you did change CT to AG at rs1408799, and GG toCC at position 8, right?

Can you doa screen shot?

is this all you need?

https://s20.postimg.org/k3sfcdtyl/skin.jpg (https://postimages.org/)

my original is

ggggctaannggaaggnnnn

firemonkey
11-27-2017, 05:54 PM
All other Europeans seem to be getting in the high 90s for white. Does this say anything about ancient ancestry?

Tolan
11-27-2017, 06:28 PM
My result (!?)
I have white skin that does not tan very well.
There are probably important genes missing.


White 9.906421 2.94%
Intermediate 9.146165 17.86%
Black 14.296864 0.00%


This profile is 2.14 times more likely Intermediate than White, and 173 times more likely Intermediate than Black.

Predicted admixture: 67.87 % for Intermediate; 31.73 % for White; 0.39 % for Black.

Therefore, the profile should be Intermediate.

CelticGerman
11-27-2017, 07:02 PM
Well it does not work at position rs1408799 (I have CT; I should put AG, right?) and rs3829241 (I have AA; shall I put CC or TT instead???).

anglesqueville
11-27-2017, 07:21 PM
Well it does not work at position rs1408799 (I have CT; I should put AG, right?) and rs3829241 (I have AA; shall I put CC or TT instead???).

Damned flipped snps. Replacement: A<->T ; C<->G

CelticGerman
11-27-2017, 07:33 PM
20012

My result. I would say 2, I am North German, but not at all Scandinavian like. I tan very quickly and I support a lot of sunshine without protection before having problems.

anglesqueville
11-27-2017, 07:33 PM
For my father:
20011

For both I would say that the accuracy is pretty good. The shade of "white" for both is the first one (but my mom is in fact paler than that).

CelticGerman
11-27-2017, 07:50 PM
French woman, Scandinavian like, tans very slowly ...20013

Kurd
11-27-2017, 08:03 PM
My result (!?)
I have white skin that does not tan very well.
There are probably important genes missing.


White 9.906421 2.94%
Intermediate 9.146165 17.86%
Black 14.296864 0.00%


This profile is 2.14 times more likely Intermediate than White, and 173 times more likely Intermediate than Black.

Predicted admixture: 67.87 % for Intermediate; 31.73 % for White; 0.39 % for Black.

Therefore, the profile should be Intermediate.

Can you post a screenshot please. The table is just as informative to me as the result


Edit: Yours seems the most off so far followed by Firemonkey

Kurd
11-27-2017, 08:05 PM
For my father:
20011

For both I would say that the accuracy is pretty good. The shade of "white" for both is the first one (but my mom is in fact paler than that).

You are fortunate that you have all the SNPs genotyped. That likely helps accuracy

firemonkey
11-27-2017, 09:12 PM
Found another snp result. Results now more in line with other Europeans.

20014

Kurd
11-27-2017, 10:11 PM
Found another snp result. Results now more in line with other Europeans.

http://www.anthrogenica.com/attachment.php?attachmentid=20014&stc=1

can you elaborate how you found it because it’s not genotyped by 23andme

firemonkey
11-27-2017, 10:20 PM
can you elaborate how you found it because it’s not genotyped by 23andme

With my FTDNA data.

Kurd
11-27-2017, 10:30 PM
With my FTDNA data.

what a difference an SNP makes. It is reasonable to assume that if Surbakhun, the Kurds and others had the missing 3, the accuracy of their results would increase, although we need more users from various continents to assess better

anglesqueville
11-27-2017, 10:41 PM
You are fortunate that you have all the SNPs genotyped. That likely helps accuracy

On each set, one snp is imputed ( don't recall which one). For my parents I work now with a a pannel of files that I got on the basis of the imputed vcf from DNA.Land, filtered through the probabilities of the genotypes. For this test I choosed the smallest files ( around 1 million snps), of maximal average proba ( >0.995 if I recall).
edit: so far I've been too lazy to make the job for myself, my own plink files are still waiting on a USB stick...

Reza
11-27-2017, 10:53 PM
Here are my family's Kurd - Pretty inaccurate in terms of detailing South Asian skin colour.

The colour palette is pretty poor as an accurate representation.

Both my parents are possibly white number 3, intermediate 1.

Untanned, I'm probably intermediate 2. Wife is intermediate 1.

Cousin is probably intermediate 2.

Or shift accordingly in either direction if I'm being colour blind.

Relative to each other it's not terribly inaccurate except wife is a bit lighter than me despite scoring more intermediate.

Mine:

https://i.imgur.com/0eyx4m0.png

Mother:

https://i.imgur.com/kmUYiGv.png

Father:

https://i.imgur.com/MpcGDw9.png

Wife:

https://i.imgur.com/YDON9g6.png

Cousin:

https://i.imgur.com/6G730AZ.png

Kurd
11-28-2017, 12:40 AM
Here are my family's Kurd - Pretty inaccurate in terms of detailing South Asian skin colour.

The colour palette is pretty poor as an accurate representation.

Both my parents are possibly white number 3, intermediate 1.

Untanned, I'm probably intermediate 2. Wife is intermediate 1.

Cousin is probably intermediate 2.

Or shift accordingly in either direction if I'm being colour blind.

Relative to each other it's not terribly inaccurate except wife is a bit lighter than me despite scoring more


Thanks for posting all of them. As I suspected their cohort most likely didn’t include any Asians. I’m working on something which I think should work much better for everyone.

Incidentally, a big application of this would be at a crime scene for ruling out black or white suspects when there isn’t any DNA info on the suspects in criminal justice databases

bmoney
11-28-2017, 01:13 AM
Thanks for posting all of them. As I suspected their cohort most likely didnít include any Asians. Iím working on something which I think should work much better for everyone.

I got a lot of NNs and one indetermined in my 23andme v5- Im definitely not white

White 1.314140 72.06%
Intermediate 9.299012 0.00%
Black 17.735324 0.00%


This profile is 2,936 times more likely White than Intermediate, and 13,540,341 times more likely White than Black.

Predicted admixture: 99.97 % for White; 0.03 % for Intermediate; 0.00 % for Black.

Therefore, the profile should be White.

Kurd
11-28-2017, 01:19 AM
I got a lot of NNs and one indetermined in my 23andme v5- Im definitely not white

White 1.314140 72.06%
Intermediate 9.299012 0.00%
Black 17.735324 0.00%


This profile is 2,936 times more likely White than Intermediate, and 13,540,341 times more likely White than Black.

Predicted admixture: 99.97 % for White; 0.03 % for Intermediate; 0.00 % for Black.

Therefore, the profile should be White.

Two things working against you:

1- Accuracy seems to drop off exponentially with fewer SNPs
2- Accuracy is lower for Asians

cps
11-28-2017, 02:08 AM
My background is Greek. Funny thing is people are always disagreeing on how dark I am. Some people say I'm olive, others say I have light skin, others say it's tan. I think my skin color is pretty light, just I have strong undertones and my face holds a tan FOREVER. It never fades to match the rest of my body.

noman
11-28-2017, 02:40 AM
I can't do. Not all mentioned markers are available in my raw data.

Tolan
11-28-2017, 09:19 AM
Can you post a screenshot please. The table is just as informative to me as the result


Edit: Yours seems the most off so far followed by Firemonkey

Here:
20026

Tolan
11-28-2017, 09:42 AM
There is a big problem with rs16891982!

See: https://www.snpedia.com/index.php/Rs16891982
GG: Generally European; Light skin; Possibly an increased risk of melanoma
CC: generally non-European, but if European, 7x more likely to have black hair

I am GG, but if i put CC, I become "white".
There must be an error on this SNP!

20027

firemonkey
11-28-2017, 10:18 AM
Just out of curiosity tried the other choices


Multinomial logistic regression algorithm.


The likelihoods of your profile are:

White Intermediate Black
1.0478e-91 1.0000e+00 6.4217e-299

Predicted admixture:
White Intermediate Black
0.00 % 100.00 % 0.00 %

This individual is more than a billion (109) times more likely to be Intermediate rather than White.

Therefore, this individual is most likely to be Intermediate.

Performing logistic regression with Intermediate as pivot population...

Loading required package: nnet
# weights: 54 (34 variable)
initial value 120.847352
iter 10 value 9.286072
iter 20 value 3.361739
iter 30 value 3.245533
iter 40 value 3.245360
final value 3.245360
converged
Intermediate Black White
1.000000e+00 6.326099e-162 1.493314e-09
> proc.time()
user system elapsed
0.156 0.021 0.169

The likelihoods of your profile are:
White Intermediate Black
1.4933e-09 1.0000e+00 6.3261e-162

Predicted admixture:
White Intermediate Black
0.00 % 100.00 % 0.00 %

This individual is 669,651,527 times more likely to be Intermediate rather than White.

Therefore, this individual is most likely to be Intermediate.

Performing logistic regression with Black as pivot population...

Loading required package: nnet
# weights: 54 (34 variable)
initial value 120.847352
iter 10 value 13.827990
iter 20 value 3.884220
iter 30 value 3.250048
iter 40 value 3.245373
final value 3.245360
converged
Black Intermediate White
0.000000e+00 1.000000e+00 3.958079e-191
> proc.time()
user system elapsed
0.153 0.020 0.167

The likelihoods of your profile are:
White Intermediate Black
3.9581e-191 1.0000e+00 0.0000e+00

Predicted admixture:
White Intermediate Black
0.00 % 100.00 % 0.00 %

This individual is more than a billion (109) times more likely to be Intermediate rather than White.

Therefore, this individual is most likely to be Intermediate.




Average classification:
White Intermediate Black
0.00 % 100.00 % 0.00 %

This individual is more than a billion (109) times more likely to be Intermediate rather than White.

Therefore, this individual is most likely to be Intermediate.

Genetic distance algorithm

Threshold genetic distance per locus D is 0.67.

Mean genetic distances per locus Dak are:
White 0.64
Intermediate 0.81
Black 0.99


Allele distances White Intermediate Black
White 0.00 0.17 0.34
Intermediate 0.17 0.00 0.18
Black 0.34 0.18 0.00

Genetic relatedness Rak for your profile is:
Black 0.32
Intermediate 0.50
White 0.67

Finally, contributions Cak sorted in descending order for your profile are:
White 54.54 % Skin colours
Intermediate 29.37 % Skin colours
Black 16.09 % Skin colours



Here are the results for the genotype version of the algorithm:


Threshold genetic distance per locus D is 0.67.

Mean genetic distances per locus Dgk are:
White 1.14
Intermediate 1.33
Black 1.44


Genotype distances White Intermediate Black
White 0.00 0.19 0.31
Intermediate 0.19 0.00 0.12
Black 0.31 0.12 0.00

Genetic relatedness Rgk for your profile is:
Black 0.36
Intermediate 0.47
White 0.67

Finally, contributions Cgk sorted in descending order for your profile are:
White 55.49 % Skin colours
Intermediate 26.62 % Skin colours
Black 17.89 % Skin colours

Kurd
11-28-2017, 12:25 PM
There is a big problem with rs16891982!

See: https://www.snpedia.com/index.php/Rs16891982
GG: Generally European; Light skin; Possibly an increased risk of melanoma
CC: generally non-European, but if European, 7x more likely to have black hair

I am GG, but if i put CC, I become "white".
There must be an error on this SNP!

http://www.anthrogenica.com/attachment.php?attachmentid=20027&stc=1

Good catch!! Seems like an error on their part. Can everyone with GG at that locus re-run with CC and repost. If anyone is CC (Asians) please redo with GG instead. Thanks

firemonkey
11-28-2017, 12:46 PM
Good catch!! Seems like an error on their part. Can everyone with GG at that locus re-run with CC and repost. If anyone is CC (Asians) please redo with GG instead. Thanks


20031

Kurd
11-28-2017, 01:04 PM
http://www.anthrogenica.com/attachment.php?attachmentid=20031&stc=1

Results are making more sense now

vettor
11-28-2017, 05:15 PM
Good catch!! Seems like an error on their part. Can everyone with GG at that locus re-run with CC and repost. If anyone is CC (Asians) please redo with GG instead. Thanks

https://s20.postimg.org/gme0dscwd/skin.jpg (https://postimg.org/image/v5l5f7615/)

mine ....now 100% white

Edward J
11-28-2017, 05:41 PM
20035

anglesqueville
11-28-2017, 06:06 PM
mom20036
dad20037

Reza
11-28-2017, 07:27 PM
Results make much more sense now with that correction for the error with rs16891982!

They actually correlate well with my self identification from the colour palette, and confirm the subtle differences that are noticeable within my own family i.e. both parents white 3 / intermediate 1, myself intermediate 2, wife intermediate 1/2, cousin intermediate 2

Myself

https://i.imgur.com/y5JY0XB.png

Mother

https://i.imgur.com/kjrrP2e.png

Father

https://i.imgur.com/okHzEbD.png

Wife

https://i.imgur.com/0c49T2l.png

Cousin

https://i.imgur.com/U0FE1Ks.png

poi
11-28-2017, 07:40 PM
I have my data from 23andme v5. I can only find 4 snps. Furthermore, only 3 snps have A/C/G and one of them has T. If I take it out the snp with T, only 3 snps are in play (out of 10), so not sure how reliable things will be.

NNNNNNAANNCGNNNNAGNN

20040

Kurd
11-28-2017, 09:11 PM
I have my data from 23andme v5. I can only find 4 snps. Furthermore, only 3 snps have A/C/G and one of them has T. If I take it out the snp with T, only 3 snps are in play (out of 10), so not sure how reliable things will be.

NNNNNNAANNCGNNNNAGNN

http://www.anthrogenica.com/attachment.php?attachmentid=20040&stc=1


Very inaccuate with only 3 or 4 snps

poi
11-28-2017, 09:31 PM
Very inaccuate with only 3 or 4 snps

23andme v5 took out 3 out of those 7 listed SNPs from v4, but still has kept their skin pigmentation prediction under "Traits". wth

Kurd
11-29-2017, 02:25 AM
Here is a summary of the posters so far. Table is sorted with highest predicted white on top. Europeans shaded green and Asians brown.
I only included users with corrected rs16891982. Again, there is a mistake which was noticed by Tolan. To correct for this users who are GG for rs16891982 need to input CC instead, and those who are CC (Asians) need to input GG. Hetrozygous users (CG) don't need to make a correction. Accuracy would have been higher if 23andMe had genotyped the missing 3 SNPs.



User
Ethnicity
% White
% Intermediate
% Black
Likelihood White to Intermediate
Likelihood White to Black
Likelihood Intermediate to White
Likelihood Intermediate to Black


Angleaqueville-mom
NW Europe
100.00%
0.00%
0.00%
7,233,373
> 1,000,000,000
N/A
N/A


Angleaqueville-dad
NW Europe
100.00%
0.00%
0.00%
908,486
> 1,000,000,000
N/A
N/A


Edward J
German/Italian
100.00%
0.00%
0.00%
1,363,733
840,086,146
N/A
N/A


Firemonkey
British
100.00%
0.00%
0.00%
40,557
> 1,000,000,000
N/A
N/A


Vettor
Italian
100.00%
0.00%
0.00%
60,823
326,329,772
N/A
N/A


Cps
Greek
99.99%
0.01%
0.00%
12,741
9,810,806
N/A
N/A


Dorkeymon
Romanian
99.99%
0.01%
0.00%
12,741
9,810,806
N/A
N/A


Kurd-Iraq-N1
Kurd
94.36%
5.64%
0.00%
17
25,312
N/A
N/A


Kurd-Iraq-N2
Kurd
89.33%
10.65%
0.02%
8
4,483
N/A
N/A


Reza-mom
Bengali
98.77%
1.21%
0.03%
82
3835
N/A
N/A


Reza-dad
Bengali
57.34%
40.70%
1.96%
1.4
29
N/A
N/A


Reza
Bengali
21.37%
74.80%
3.82%
N/A
N/A
3.5
19.6


Kurd
Kurd/Baloch
14.03%
82.17%
3.80%
N/A
N/A
5.9
21.6


Reza-cousin
Bengali
2.01%
86.65%
11.34%
N/A
N/A
43.1
7.6





EDIT: Users who did not update their rs16891982 were not added to table. Also, 23 V5 users with only 3 or 4 overlapping SNPs were not included, because the margin of error is too high

parasar
11-29-2017, 03:02 AM
Something looks off as I have generally used just one SNP - rs16891982 (http://www.ias.ac.in/article/fulltext/jgen/092/01/0135-0139) - for western Eurasians, and that puts me midway accurately (perceived).

FWIW now with seven,
White 4.462646 95.59%
Intermediate 14.355624 0.00%
Black 22.424732 0.00%

This profile is 19,791 times more likely White than Intermediate, and 63,217,138 times more likely White than Black.

Predicted admixture: 99.99 % for White; 0.01 % for Intermediate; 0.00 % for Black.

Therefore, the profile should be White.

Kurd
11-29-2017, 03:12 AM
Something looks off as I have generally used just one SNP - rs16891982 (http://www.ias.ac.in/article/fulltext/jgen/092/01/0135-0139) - for western Eurasians, and that puts me midway accurately (perceived).

FWIW now with seven,
White 4.462646 95.59%
Intermediate 14.355624 0.00%
Black 22.424732 0.00%

This profile is 19,791 times more likely White than Intermediate, and 63,217,138 times more likely White than Black.

Predicted admixture: 99.99 % for White; 0.01 % for Intermediate; 0.00 % for Black.

Therefore, the profile should be White.


What’s your genotype for that SNP; If you’re CC you have to input GG

parasar
11-29-2017, 03:15 AM
1- What’s your genotype for that SNP;
2- what’s your result when you input seven or more SNPs?

1. CG.
2. The one I posted (only 7 available from 23andme)

Kurd
11-29-2017, 03:19 AM
1. CG.
2. The one I posted (only 7 available from 23andme)

I agree, it seems off for some Asians. It is likely that the missing 3 SNPs are a factor. Also, I believe accuracy would have been higher if 30 SNPs were used. I’m working on something that uses 36 SNPs

Kurd
11-29-2017, 03:46 AM
Added Sangarius. He's CG, no update necessary



User
Ethnicity
% White
% Intermediate
% Black
Likelihood White to Intermediate
Likelihood White to Black
Likelihood Intermediate to White
Likelihood Intermediate to Black


Angleaqueville-mom
NW Europe
100.00%
0.00%
0.00%
7,233,373
> 1,000,000,000
N/A
N/A


Angleaqueville-dad
NW Europe
100.00%
0.00%
0.00%
908,486
> 1,000,000,000
N/A
N/A


Edward J
German/Italian
100.00%
0.00%
0.00%
1,363,733
840,086,146
N/A
N/A


Firemonkey
British
100.00%
0.00%
0.00%
40,557
> 1,000,000,000
N/A
N/A


Vettor
Italian
100.00%
0.00%
0.00%
60,823
326,329,772
N/A
N/A


Cps
Greek
99.99%
0.01%
0.00%
12,741
9,810,806
N/A
N/A


Dorkeymon
Romanian
99.99%
0.01%
0.00%
12,741
9,810,806
N/A
N/A


Reza-mom
Bengali
98.77%
1.21%
0.03%
82
3835
N/A
N/A


Sangarius
Turkish
98.55%
1.44%
0.01%
68
16,846
N/A
N/A


Kurd-Iraq-N1
Kurd
94.36%
5.64%
0.00%
17
25,312
N/A
N/A


Kurd-Iraq-N2
Kurd
89.33%
10.65%
0.02%
8
4,483
N/A
N/A


Reza-dad
Bengali
57.34%
40.70%
1.96%
1.4
29
N/A
N/A


Reza
Bengali
21.37%
74.80%
3.82%
N/A
N/A
3.5
19.6


Kurd
Kurd/Baloch
14.03%
82.17%
3.80%
N/A
N/A
5.9
21.6


Reza-cousin
Bengali
2.01%
86.65%
11.34%
N/A
N/A
43.1
7.6

vettor
11-29-2017, 05:09 AM
my father Lucio results ..........he is darker than me

https://s20.postimg.org/gqzuo7uv1/lucio_skin.jpg (https://postimg.org/image/qbjhb3k6x/)

Sangarius
11-29-2017, 10:14 AM
Added Sangarius. He's CG, no update necessary



User
Ethnicity
% White
% Intermediate
% Black
Likelihood White to Intermediate
Likelihood White to Black
Likelihood Intermediate to White
Likelihood Intermediate to Black


Angleaqueville-mom
NW Europe
100.00%
0.00%
0.00%
7,233,373
> 1,000,000,000
N/A
N/A


Angleaqueville-dad
NW Europe
100.00%
0.00%
0.00%
908,486
> 1,000,000,000
N/A
N/A


Edward J
German/Italian
100.00%
0.00%
0.00%
1,363,733
840,086,146
N/A
N/A


Firemonkey
British
100.00%
0.00%
0.00%
40,557
> 1,000,000,000
N/A
N/A


Vettor
Italian
100.00%
0.00%
0.00%
60,823
326,329,772
N/A
N/A


Cps
Greek
99.99%
0.01%
0.00%
12,741
9,810,806
N/A
N/A


Dorkeymon
Romanian
99.99%
0.01%
0.00%
12,741
9,810,806
N/A
N/A


Reza-mom
Bengali
98.77%
1.21%
0.03%
82
3835
N/A
N/A


Sangarius
Turkish
98.55%
1.44%
0.01%
68
16,846
N/A
N/A


Kurd-Iraq-N1
Kurd
94.36%
5.64%
0.00%
17
25,312
N/A
N/A


Kurd-Iraq-N2
Kurd
89.33%
10.65%
0.02%
8
4,483
N/A
N/A


Reza-dad
Bengali
57.34%
40.70%
1.96%
1.4
29
N/A
N/A


Reza
Bengali
21.37%
74.80%
3.82%
N/A
N/A
3.5
19.6


Kurd
Kurd/Baloch
14.03%
82.17%
3.80%
N/A
N/A
5.9
21.6


Reza-cousin
Bengali
2.01%
86.65%
11.34%
N/A
N/A
43.1
7.6



I have noticed something odd. Shouldn't Reza cluster between his parents? Instead he is predicted to be even darker than his dad. I assume Reza might have made a mistake while entering the allelles of his own or his mother's. Given their ancestry (Bengali) I suspect that it is the the latter.

Reza
11-29-2017, 10:29 AM
I have noticed something odd. Shouldn't Reza cluster between his parents? Instead he is predicted to be even darker than his dad. I assume Reza might have made a mistake while entering the allelles of his own or his mother's. Given their ancestry (Bengali) I suspect that it is the the latter.

No, I rechecked, they're accurate.

My parents are heterozygous on various snps and I seem to have inherited the darker variation on many.

Which correlates phenotypically, as I have the darkest pigmentation compared to my siblings.

The above few snps clearly don't explain all the variations in South Asian skin colour as I'm sure my mother isn't as light as Sangarius, or your average Kurd.

As Kurd mentioned, it's probably more useful to compare between similar ethnic groups?

Sangarius
11-29-2017, 10:42 AM
No, I rechecked, they're accurate.

My parents are heterozygous on various snps and I seem to have inherited the darker variation on many.

Which correlates phenotypically, as I have the darkest pigmentation compared to my siblings.

The above few snps clearly don't explain all the variations in South Asian skin colour as I'm sure my mother isn't as light as Sangarius, or your average Kurd.

As Kurd mentioned, it's probably more useful to compare between similar ethnic groups?

I see, that's another possible explanation that I considered, but deemed unlikely. Interesting how that goes sometimes, my father is also darker than either of his parents.

Reza
11-29-2017, 11:27 AM
I see, that's another possible explanation that I considered, but deemed unlikely. Interesting how that goes sometimes, my father is also darker than either of his parents.

Yeah, I'm darker than both my parents, who are also lighter than most of their siblings.

From limited experience, I've noted there's a lot of variation with pigmentation amongst Turkish populations. A result of being quite an admixed population?

Kurd
11-29-2017, 01:01 PM
Updated with Lucio:



User
Ethnicity
% White
% Intermediate
% Black
Likelihood White to Intermediate
Likelihood White to Black
Likelihood Intermediate to White
Likelihood Intermediate to Black


Angleaqueville-mom
NW Europe
100.00%
0.00%
0.00%
7,233,373
> 1,000,000,000
N/A
N/A


Edward J
German/Italian
100.00%
0.00%
0.00%
1,363,733
840,086,146
N/A
N/A


Angleaqueville-dad
NW Europe
100.00%
0.00%
0.00%
908,486
> 1,000,000,000
N/A
N/A


Firemonkey
British
100.00%
0.00%
0.00%
40,557
> 1,000,000,000
N/A
N/A


Vettor
Italian
100.00%
0.00%
0.00%
60,823
326,329,772
N/A
N/A


Lucio
Italian
99.99%
0.01%
0.00%
12,741
9,810,806
N/A
N/A


Cps
Greek
99.99%
0.01%
0.00%
12,741
9,810,806
N/A
N/A


Dorkeymon
Romanian
99.99%
0.01%
0.00%
12,741
9,810,806
N/A
N/A


Reza-mom
Bengali
98.77%
1.21%
0.03%
82
3835
N/A
N/A


Sangarius
Turkish
98.55%
1.44%
0.01%
68
16,846
N/A
N/A


Kurd-Iraq-N1
Kurd
94.36%
5.64%
0.00%
17
25,312
N/A
N/A


Kurd-Iraq-N2
Kurd
89.33%
10.65%
0.02%
8
4,483
N/A
N/A


Reza-dad
Bengali
57.34%
40.70%
1.96%
1.4
29
N/A
N/A


Reza
Bengali
21.37%
74.80%
3.82%
N/A
N/A
3.5
19.6


Kurd
Kurd/Baloch
14.03%
82.17%
3.80%
N/A
N/A
5.9
21.6


Reza-cousin
Bengali
2.01%
86.65%
11.34%
N/A
N/A
43.1
7.6

Kurd
11-29-2017, 01:09 PM
No, I rechecked, they're accurate.

My parents are heterozygous on various snps and I seem to have inherited the darker variation on many.

Which correlates phenotypically, as I have the darkest pigmentation compared to my siblings.

The above few snps clearly don't explain all the variations in South Asian skin colour as I'm sure my mother isn't as light as Sangarius, or your average Kurd.

As Kurd mentioned, it's probably more useful to compare between similar ethnic groups?

I am curious, how would you describe the natural skin color of your mom with hair and eye color. Kurd-Iraq-North2 is average for Kurds from that area. He is between White3 and Intermediate1 (olive) untanned, black hair, brown eyes.

Ignis90
11-29-2017, 01:26 PM
23andme v3, 3 missing SNPs.

20048

Kurd
11-29-2017, 01:33 PM
I am curious, how would you describe the natural skin color of your mom with hair and eye color. Kurd-Iraq-North2 is average for Kurds from that area. He is between White3 and Intermediate1 (olive) untanned, black hair, brown eyes.


In the interest of communicating the skin color for Kurd-Iraq-N2, I am posting a photo of him. He is average complexion for males from his area in N Iraq. This is one taken in the summer, so it would be reasonable to assume that he is tanned. I don't have a winter photo for him. He is also one of the Kurd C samples I use for analysis. Clusters with Iraqi Kurds and Iranians.


20049

Kurd
11-29-2017, 01:35 PM
23andme v3, 3 missing SNPs.

http://www.anthrogenica.com/attachment.php?attachmentid=20048&stc=1

What ethnicity?

Ignis90
11-29-2017, 02:24 PM
What ethnicity?
Berber.

Kurd
11-30-2017, 12:50 AM
Updated with Ignis90 and Varun




User
Ethnicity
% White
% Intermediate
% Black
Likelihood White to Intermediate
Likelihood White to Black
Likelihood Intermediate to White
Likelihood Intermediate to Black


Angleaqueville-mom
NW Europe
100.00%
0.00%
0.00%
7,233,373
> 1,000,000,000
N/A
N/A


Edward J
German/Italian
100.00%
0.00%
0.00%
1,363,733
840,086,146
N/A
N/A


Angleaqueville-dad
NW Europe
100.00%
0.00%
0.00%
908,486
> 1,000,000,000
N/A
N/A


Firemonkey
British
100.00%
0.00%
0.00%
40,557
> 1,000,000,000
N/A
N/A


Vettor
Italian
100.00%
0.00%
0.00%
60,823
326,329,772
N/A
N/A


Lucio
Italian
99.99%
0.01%
0.00%
12,741
9,810,806
N/A
N/A


Cps
Greek
99.99%
0.01%
0.00%
12,741
9,810,806
N/A
N/A


Dorkeymon
Romanian
99.99%
0.01%
0.00%
12,741
9,810,806
N/A
N/A


Ignis90
Berber
99.94%
0.06%
0.00%
1,555
4,917,319
N/A
N/A


Reza-mom
Bengali
98.77%
1.21%
0.03%
82
3835
N/A
N/A


Sangarius
Turkish
98.55%
1.44%
0.01%
68
16,846
N/A
N/A


Kurd-Iraq-N1
Kurd
94.36%
5.64%
0.00%
17
25,312
N/A
N/A


Kurd-Iraq-N2
Kurd
89.33%
10.65%
0.02%
8
4,483
N/A
N/A


Reza-dad
Bengali
57.34%
40.70%
1.96%
1.4
29
N/A
N/A


Reza
Bengali
21.37%
74.80%
3.82%
N/A
N/A
3.5
19.6


Kurd
Kurd/Baloch
14.03%
82.17%
3.80%
N/A
N/A
5.9
21.6


Reza-cousin
Bengali
2.01%
86.65%
11.34%
N/A
N/A
43.1
7.6


Varun
S Indian Brahmin
5.64%
93.85%
0.52%
N/A
N/A
16.64
182

Varun R
11-30-2017, 02:03 AM
My results

20057

Intermediate 3 is quite close to my skin color.

surbakhunWeesste
11-30-2017, 02:41 AM
Good catch!! Seems like an error on their part. Can everyone with GG at that locus re-run with CC and repost. If anyone is CC (Asians) please redo with GG instead. Thanks

well, it still holds correct for my dad,
http://www.anthrogenica.com/showthread.php?12782-Online-Pigmentation-Predictor&p=316206&viewfull=1#post316206

for mine I may have done it wrong cz I can get really dyslexic sometimes, IRRC this correlates to DMXX's eye chart project based on the snps. I took for ever to do that but did it correctly, but i will try again later, even though it ain't gonna be accurate for me.

edit: I did mine incorrect.
http://www.anthrogenica.com/showthread.php?1057-DMXX-s-Eye-Colour-Project-(v-2)/page14

Kurd
11-30-2017, 12:37 PM
well, it still holds correct for my dad,
http://www.anthrogenica.com/showthread.php?12782-Online-Pigmentation-Predictor&p=316206&viewfull=1#post316206

for mine I may have done it wrong cz I can get really dyslexic sometimes, IRRC this correlates to DMXX's eye chart project based on the snps. I took for ever to do that but did it correctly, but i will try again later, even though it ain't gonna be accurate for me.

edit: I did mine incorrect.
http://www.anthrogenica.com/showthread.php?1057-DMXX-s-Eye-Colour-Project-(v-2)/page14



Wow, no wonder your and your dad's results were so off.... BTW, that link was no good, I found your results at http://www.anthrogenica.com/showthread.php?1057-DMXX-s-Eye-Colour-Project-(v-2)/page20


I summarized for you what you entered differently in DMXX's Eye Color project versus here. I have no idea which ones are correct. Please redo yours and your dad's whenever you have time and repost. Let me know if you need help. Hopefully, they will be more in line with Asians :)



SNPs
Surbakhun-DMXX Project
Surbakhun-Pigment Project


rs10777129
N/A
NN


rs13289
N/A
CC


rs1408799
TT
AG


rs1426654
N/A
AA


rs1448484
N/A
NN


rs16891982
CG
CC


rs2402130
N/A
AG


rs3829241
CC
NN


rs6058017
N/A
GG


rs6119471
N/A
NN






SNPs
Surbakhun DAD-DMXX Project
Surbakhun DAD-Pigment Project


rs10777129
N/A
NN


rs13289
N/A
CC


rs1408799
TT
GG


rs1426654
N/A
AA


rs1448484
N/A
NN


rs16891982
CG
CG


rs2402130
N/A
AA


rs3829241
CC
CT


rs6058017
N/A
AA


rs6119471
N/A
NN

surbakhunWeesste
11-30-2017, 07:14 PM
Wow, no wonder your and your dad's results were so off.... BTW, that link was no good, I found your results at http://www.anthrogenica.com/showthread.php?1057-DMXX-s-Eye-Colour-Project-(v-2)/page20


I summarized for you what you entered differently in DMXX's Eye Color project versus here. I have no idea which ones are correct. Please redo yours and your dad's whenever you have time and repost. Let me know if you need help. Hopefully, they will be more in line with Asians :)



SNPs
Surbakhun-DMXX Project
Surbakhun-Pigment Project


rs10777129
N/A
NN


rs13289
N/A
CC


rs1408799
TT
AG


rs1426654
N/A
AA


rs1448484
N/A
NN


rs16891982
CG
CC


rs2402130
N/A
AG


rs3829241
CC
NN


rs6058017
N/A
GG


rs6119471
N/A
NN






SNPs
Surbakhun DAD-DMXX Project
Surbakhun DAD-Pigment Project


rs10777129
N/A
NN


rs13289
N/A
CC


rs1408799
TT
GG


rs1426654
N/A
AA


rs1448484
N/A
NN


rs16891982
CG
CG


rs2402130
N/A
AA


rs3829241
CC
CT


rs6058017
N/A
AA


rs6119471
N/A
NN


You have his raw data anyway, lol you can cross check with the snps if necessary. I might do it later, thanks :)

MonkeyDLuffy
12-01-2017, 02:58 AM
Mine:

I'm close to the third one, not sure if I did it right though:

https://i.imgur.com/nW0m8oz.png

Kurd
12-01-2017, 03:42 AM
Mine:

I'm close to the third one, not sure if I did it right though:

https://i.imgur.com/nW0m8oz.png


You need to change Rs16891982 to GG (their assignment is incorrect). Once you do that your prediction will become darker

MonkeyDLuffy
12-01-2017, 04:11 AM
You need to change Rs16891982 to GG (their assignment is incorrect). Once you do that your prediction will become darker

Here's the newer results:

https://i.imgur.com/ey5FCf7.png

Kurd
12-01-2017, 04:22 AM
Here's the newer results:

https://i.imgur.com/ey5FCf7.png



Better but still off for you. Clearly there are other SNPs associated with pigmentation in S Asians than the 7 here

MonkeyDLuffy
12-01-2017, 04:32 AM
Better but still off for you. Clearly there are other SNPs associated with pigmentation in S Asians than the 7 here

I guess, I’m pretty light for Punjabi standards. Kind of runs in the family.

Kurd
12-01-2017, 01:12 PM
I guess, Iím pretty light for Punjabi standards. Kind of runs in the family.

Itís still off for you. Hopefully, Iíll have something based on 36 SNPs that works better for S Asians

poi
12-01-2017, 05:34 PM
It’s still off for you. Hopefully, I’ll have something based on 36 SNPs that works better for S Asians

FWIW, my 23andme v5 'traits' page proudly shows the Skin Pigmentation prediction based on -- "two genetic markers", lol. To be fair, they do say(disclaim) that the skin pigmentation works best for "Europeans and Africans".

surbakhunWeesste
12-01-2017, 08:33 PM
this is what my 23&me says, they're never accurate in my case


20138


edit: Checked my relatives profiles, they all show same as mine lulz. We have been blanketed by 23&me?

MonkeyDLuffy
12-01-2017, 10:35 PM
https://i.imgur.com/nnI6Kwy.png

Light beige seems right for me.

jesus
12-01-2017, 10:54 PM
23andme uses 4 SNPs only in their skin tone tool. Quite pathetic for a multi million dollar company.

poi
12-01-2017, 10:59 PM
23andme uses 4 SNPs only in their skin tone tool. Quite pathetic for a multi million dollar company.

They disclaim is by saying it works best with Europeans and Africans. Also, they can't be multi million by spending resources on those SNPs, you know

if this is 23andme time, here is mine lol

20142

surbakhunWeesste
12-01-2017, 11:05 PM
They disclaim is by saying it works best with Europeans and Africans. Also, they can't be multi million by spending resources on those SNPs, you know

if this is 23andme time, here is mine lol

20142
They are a research company and they should be using all that they could, even independent “researchers” are trying to create more detailed version. They suck.

There is no way you and I are of the same complexion, lulz. See how blanketed it looks, and I am pretty sure it would be same for European or African folks. They just gave a spectrum and asked us to suck it up, in the same manner how they give us the ancestry breakdown.

jesus
12-01-2017, 11:10 PM
They disclaim is by saying it works best with Europeans and Africans. Also, they can't be multi million by spending resources on those SNPs, you know

if this is 23andme time, here is mine lol

20142

Even independent individuals like Kurd are using more SNPs in their pigmentation tools(for free). There are many research papers regarding this subject(and most of them are free to the public) with countless phenotype related SNPs. This applies to many of their tools as well. They're quite lazy in all honesty.

Reza
12-01-2017, 11:12 PM
They are a research company and they should be using all that they could, even independent ďresearchersĒ are trying to create more detailed version. They suck.

There is no way you and I are of the same complexion, lulz. See how blanketed it looks, and I am pretty sure it would be same for European or African folks. They just gave a spectrum and asked us to suck it up, in the same manner how they give us the ancestry breakdown.

I have the same breakdown as MonkeyDLuffy and I'm definitely much darker than he is.

It is frustrating that, at least in the UK, you spend £149 on the kit, and get very little back for it - a generic South Asian which encompasses 1.5 billion people with no further breakdown, and health traits, which are mostly meaningless or over simplified. It's only for the raw data that I've spent so much on it.

poi
12-01-2017, 11:14 PM
I have the same breakdown as MonkeyDLuffy and I'm definitely much darker than he is.

It is frustrating that, at least in the UK, you spend £149 on the kit, and get very little back for it - a generic South Asian which encompasses 1.5 billion people with no further breakdown, and health traits, which are mostly meaningless or over simplified. It's only for the raw data that I've spent so much on it.

Their skin related SNPs are very limited. You would think they would do those for checking cancer etc. I paid for the 'health' kit especially for markers that might indicate things. They are clearly ripping off the customers.

Reza
12-01-2017, 11:24 PM
Their skin related SNPs are very limited. You would think they would do those for checking cancer etc. I paid for the 'health' kit especially for markers that might indicate things. They are clearly ripping off the customers.

The danger with cancer related health reports is that the lack of appropriate genetic counselling. Even with BRCA-1 and 2 genes, the presence of it doesn't correlate with 100pc certainty of breast cancer (albeit extremely high), and more importantly, the absence of it doesn't exclude other unknown familial breast cancers and should not provide false reassurance.

One could argue that with sufficient explanation, it's the customers perogative to know and do what they will with the report, but there are major ethical questions about marketing for profit such reports on a population level without providing the infrastructure to deal with the uncertainties and further questions they will inevitably raise.

khanabadoshi
12-01-2017, 11:54 PM
Am I the only person who got likely to have darker skin... by a massive margin I might add? LOLLL.

https://i.gyazo.com/0e98c13c8c0c11ff1007940439f0e8f7.png

Reza
12-02-2017, 12:06 AM
Am I the only person who got likely to have darker skin... by a massive margin I might add? LOLLL.

https://i.gyazo.com/0e98c13c8c0c11ff1007940439f0e8f7.png

Hah, my wife and cousin scored the same.

Presumably AG at rs1426654?

khanabadoshi
12-02-2017, 12:13 AM
Hah, my wife and cousin scored the same.

Presumably AG at rs1426654?

I didn't go thru my SNPs. It seems like such a chore. But might as well. What is the list everyone is supposed to be posting?

I just checked, yup
A / G.

khanabadoshi
12-02-2017, 12:48 AM
khanabadoshi
ALLELE
REVERSE



rs10777129
GG




rs13289
GG




rs1408799
CT
Ľ
AG


rs1426654
AG




rs1448484
NN
Ľ
NN


rs16891982
CC
Ľ
GG


rs2402130
AA




rs3829241
GG
Ľ
CC


rs6058017
AG




rs6119471
NN





In case people messed up, I bolded the 4 SNPs that are reversed. It's shown in the diagram labeled as "reverse"; for all 4 of those you're supposed to enter the opposite.


https://i.gyazo.com/d305652b14d8409a0d60df3a05a083d1.png

https://i.gyazo.com/c41fd1316cba53aba711c0f86d404c06.pnghttps://i.gyazo.com/9a0663b8e78a55b1a31e441d0fa62a6c.pnghttps://i.gyazo.com/89531d1d87122792aff66fbc9140ae9d.pnghttps://i.gyazo.com/015add75d4b3ad5c6d2c5f101199a69f.pnghttps://i.gyazo.com/617619ec86f6ac0cb3d79d316e8ede7b.pnghttps://i.gyazo.com/5c623106280dd6090b76b85d6504e14d.png

Actual photos: Skin of face, chest, arm, finger, leg, foot. I don't remember which order.
Looks pretty intermediate to me.

Kurd
12-02-2017, 04:13 PM
Even independent individuals like Kurd are using more SNPs in their pigmentation tools(for free). There are many research papers regarding this subject(and most of them are free to the public) with countless phenotype related SNPs. This applies to many of their tools as well. They're quite lazy in all honesty.

Th odd thing is that they are relying on what people put down in their profile more than the SNPs to make the prediction. For kicks, Ichanged the ethnicity for one of the Kurds from ME to SA, all of a sudden their prediction went from lighter to darker lol.

Try it yourself

poi
12-02-2017, 04:17 PM
Th odd thing is that they are relying on what people put down in their profile more than the SNPs to make the prediction. For kicks, Ichanged the ethnicity for one of the Kurds from ME to SA, all of a sudden their prediction went from lighter to darker lol.

Try it yourself

Lol, very scientific.

MonkeyDLuffy
12-02-2017, 04:51 PM
On 23andme it says "Likely to be light skin" lol. I'm no way lighter than khana, probably 1 or 2 shades darker than him.

surbakhunWeesste
12-02-2017, 06:39 PM
On 23andme it says "Likely to be light skin" lol. I'm no way lighter than khana, probably 1 or 2 shades darker than him.

The whole "light to dark" slightly light/dark also depends on your undertone, you need to check foundation shades to understand how crazy shenanigan this is.

poi
12-03-2017, 01:23 AM
The whole "light to dark" slightly light/dark also depends on your undertone, you need to check foundation shades to understand how crazy shenanigan this is.

It's difficult to get a consistent skin complexion across different subjects physically at a different location. I'm an (enthusiast) portrait photographer and it a major pain to get the complexion just right. For some demanding clients, they can be a bit picky and sensitive. But different textures reflect light differently.

surbakhunWeesste
12-03-2017, 02:01 AM
It's difficult to get a consistent skin complexion across different subjects physically at a different location. I'm an (enthusiast) portrait photographer and it a major pain to get the complexion just right. For some demanding clients, they can be a bit picky and sensitive. But different textures reflect light differently.

Talking about pics: I hate posing for camera; I can never make a straight face; I missed most of my pic days in school. I love colors but certain colors flush me out. 23&me trait section ain't helping neither is Sephora's color IQ lulz.

poi
12-03-2017, 03:41 PM
Talking about pics: I hate posing for camera; I can never make a straight face; I missed most of my pic days in school. I love colors but certain colors flush me out. 23&me trait section ain't helping neither is Sephora's color IQ lulz.

Not sure if this is relevant, and definitely going way off topic, but if you aren't sure what best suits your looks, chances are that you have certain expectations but the lighting isn't helping you achieve you that look, rather than coloring. I would just do black and white anyway for portraits. Simple household lamp and smartphone should be all you need unless you want something pro.

Typically, narrow faces need to have lighting that makes your face wider, while wide faced people need to shape lighting to make them look narrow faced. The key is the harmonic mean: neither too wide, not too long. Unless, of course , you do want to challenge the norms.

https://digital-photography-school.com/24-portrait-lighting-setups/

MonkeyDLuffy
12-03-2017, 04:28 PM
Not sure if this is relevant, and definitely going way off topic, but if you aren't sure what best suits your looks, chances are that you have certain expectations but the lighting isn't helping you achieve you that look, rather than coloring. I would just do black and white anyway for portraits. Simple household lamp and smartphone should be all you need unless you want something pro.

Typically, narrow faces need to have lighting that makes your face wider, while wide faced people need to shape lighting to make them look narrow faced. The key is the harmonic mean: neither too wide, not too long. Unless, of course , you do want to challenge the norms.

https://digital-photography-school.com/24-portrait-lighting-setups/

Also the size of lens affects face shape as well, if it is a close range lense like 16mm-24mm then your face looks narrow with bigger nose, and if it is long range like 70mm+ then your face looks more wide.

poi
12-03-2017, 05:41 PM
Also the size of lens affects face shape as well, if it is a close range lense like 16mm-24mm then your face looks narrow with bigger nose, and if it is long range like 70mm+ then your face looks more wide.

Absolutely ... the smartphone lenses are pretty damn wide for portraits if we are going for the "normal" angle of view. That is why most smartphone portraits look so unflattering... nose becomes more prominent while face becomes longer. A more pleasing shot would be from the dslr with a long focal length for sure. But new Apple and Samsung phones are getting better for portraits. They take normal angle of view with flatter faces and fake but good background blur.

khanabadoshi
12-04-2017, 06:41 AM
Corrected. See my post here for explanation --- http://www.anthrogenica.com/showthread.php?3433-Waves-of-migration-into-South-Asia&p=320103&viewfull=1#post320103

SUMMARY. If you have 23andme data, reverse your call @ rs13289 as well.

I jumped the gun :D

Kurd
12-04-2017, 02:34 PM
No need to flip rs13289 for 23 data. See explanation at thread referenced by Khana

vettor
12-05-2017, 05:31 PM
my father Lucio results ..........he is darker than me

https://s20.postimg.org/gqzuo7uv1/lucio_skin.jpg (https://postimg.org/image/qbjhb3k6x/)

I forgot to mention , if this makes a difference, I am unsure...........but my 23andme was version 3 and my father was version 4

Kurd
12-05-2017, 07:21 PM
I forgot to mention , if this makes a difference, I am unsure...........but my 23andme was version 3 and my father was version 4


the most accurate results are for those who have all 10 SNPs, and the most comparable results are using the same chipset. V4 with V4, V3 with V3, etc, because different SNPs have different probability ranges.

bmoney
12-06-2017, 03:30 AM
Absolutely ... the smartphone lenses are pretty damn wide for portraits if we are going for the "normal" angle of view. That is why most smartphone portraits look so unflattering... nose becomes more prominent while face becomes longer. A more pleasing shot would be from the dslr with a long focal length for sure. But new Apple and Samsung phones are getting better for portraits. They take normal angle of view with flatter faces and fake but good background blur.

Ive noticed this

Sangarius
12-06-2017, 02:53 PM
Update: In order to fill-in the missing markers I downloaded dna.land's imputed VCF files. I did this for my own and my grandparents kits.

Some notes on dna.land's imputed VCF files:
I hadn't bothered with dna.land's imputed VCF until now, as it seemed kinda daunting and mysterious to me.
But now that I engaged a little bit with the topic of imputation, I believe this to be the single best service offering for genomics enthusiasts out there. Either I'm missing something, or this really is a highly underappreciated service offering.

Compared to all other "genomics service platforms" that are popping up left and right (decoying users with shitty admixture runs in return for their valuable kits), users really get something of incredible value that they don't get anywhere else. About 40 million imputed SNPs with an accuracy of 99.994% (http://compass.dna.land/faq.html#Q16). While not optimal for mendelian traits, this is excellent for the really interesting highly polygenic traits concerning health, cognition, physique etc. that are targeted by GWAS studies. For example the 36 SNP skin color predictor that Kurd is working on (based on Walsh et al. 2017 (https://link.springer.com/content/pdf/10.1007%2Fs00439-017-1808-5.pdf) ?) is exactly the kind of model that the imputed VCF file is perfectly suited for. The more SNPs used by a model the less of a concern the imperfect accuracy becomes.

While I had uploaded my own and my maternal grandparents kits already in 2015, I only uploaded my paternal grandparents kits a couple days ago. Thereby I noticed a couple of things:
- the imputed VCF files of my paternal grandparents don't have the ending ".vcf.gz", but only ".vcf". But they are not VCF files, if you try to open them with glogg you will get corrupted output. You actually have to add ".gz" to the file ending. Then it's possible to unpack them or use with dna.land's compass.
- my paternal grandparents files are about 0.5gb smaller than the our VCF files and show missing markers when using with dna.land's compass.
- when opening the new VCF files with glogg, the format looks different than the old files. While quite a lot smaller in size, the new VCF files have 6 million (!) more SNPs (43.8M vs 37.8M SNPs).
- The reason: my paternal grandparents VCF files are in the newer VCF v4.1 format, while our older VCF files are in v4.0 format.
I wonder how new this change to v4.1 is, considering that dna.land hasn`t updated their website nor their compass application. I had to do quite a bit of google research to understand how to interpret the new imputation probabilities. (v4.0 format: 0/0:0.00,-10.00,-10.00 ; v4.1 format: 0|0:0 )


Pigmentation results:
My maternal grandma comes out as lightest, closely followed by my maternal grandpa. After a large gap my paternal grandma and grandpa follow. This doesn't really map onto reality, though. While all of my grandparents are definitely white in terms of coloration, the order isn't correct. It seems like the rather high East-Eurasian admixture that my paternal grandparents have (>15%) influences their results. The calculator is primed on white vs black and doesn't consider (North)-East-Eurasian variation. I wonder how a light East-Eurasian individual would score on this calculator. Anyway, looking forward to Kurd's 36 SNP calculator.


Mine (23andme v3)
https://abload.de/img/pigmentationk10-alpercqob1.png



MGM (FTDNA)
https://abload.de/img/pigmentationk10-annea3orse.png



MGP (FTDNA)
https://abload.de/img/pigmentationk10-mustaalogc.png



PGM (FTDNA)
https://abload.de/img/pigmentationk10-babaazoqs7.png



PGP (FTDNA)
https://abload.de/img/pigmentationk10-nuridybrso.png

wilhelmhalys
02-08-2018, 05:15 PM
I used the DNA.LAND imputed data to complete the 23andme V5 file SNPs that I got, and I carried out the test.
I had to change two SNPs alleles to the ones proposed by the tool.
I think is pretty accurated, even using 30% SNPs from imputed data.

21305

21306

poi
02-19-2018, 10:15 PM
Can anyone make sure I did this correctly (using my FTDNA.b37 and 23andme.V5 data):



SNP-Indels ftdna v5 combo final
rs10777129 GG - GG GG
rs13289 - - - NN
rs1408799 - TT TT GG
rs1426654 AA AA AA AA
rs1448484 - - - NN
rs16891982 - CG CG CG
rs2402130 AA - AA AA
rs3829241 AG - AG CT
rs6058017 - - - NN
rs6119471 - - - NN


Did I do this right? I am seeing weird results, so want to confirm before using imputed alleles to fill the NNs.

jeanL
02-20-2018, 02:08 PM
Let me go ahead an make a Public Service Announcement PSA:

In the above calculator rs16891982 is backwards (notice it says reverse in the website). That means that if you are GG you should enter CC; notice that the allele frequency in the parent populations are backwards with C being solely found in 84% of White; and G being found in the majority of Intermediate and all(95%) of the Black. Go ahead a correct this to get appropriate results. If you are heterozygous then is ok; but be mindful of the allele frequencies.

Aldric
02-23-2018, 03:14 AM
https://s13.postimg.org/ba12jo00n/Untitled.jpg

bmoney
02-23-2018, 04:45 AM
Can anyone make sure I did this correctly (using my FTDNA.b37 and 23andme.V5 data):



SNP-Indels ftdna v5 combo final
rs10777129 GG - GG GG
rs13289 - - - NN
rs1408799 - TT TT GG
rs1426654 AA AA AA AA
rs1448484 - - - NN
rs16891982 - CG CG CG
rs2402130 AA - AA AA
rs3829241 AG - AG CT
rs6058017 - - - NN
rs6119471 - - - NN


Did I do this right? I am seeing weird results, so want to confirm before using imputed alleles to fill the NNs.

Yo poi could you run mine too

Volat
02-23-2018, 06:10 AM
Don't know how to use the 'robust pigmentation predictor'. My results from 23andme and Gedmatch


23andme : Likely blue
GG : In Europeans, 72% chance of blue eyes; 27% chance of green eyes; 1% chance of brown eyes



rs12913832 GG
rs1800407 CC
rs12896399 GG
rs16891482 CC
rs1393350 AG
rs12203592 CC
rs7495174 AA
rs7183877 CC
rs4778232 CT
rs1408799 CT
rs8024968 CT
rs683 AC
rs4778138 AA
rs4778241 CC
rs1667394 TT
rs26722 CC
rs1015362 CT
rs4911414 GT
rs1540771 TT
rs12821256 TT
rs885478 GG
rs1805008 CT
rs1805005 GG
rs3829241 AA



Gedmatch




CT at: rs17762363 - Increased melanin production. Adds yellow, amber, or brown. Some darkening. Contributes to brown.
GG at: rs7174027 - Blocks some melanin. Often gives light colored eyes.
CC at: rs4778241 - Low Melanin. Basis for Gray, Blue, Green, or Yellow Eyes if no other pigmentation is present.
CC at: rs9782955 - Blocks some melanin. Often gives light colored eyes.
CT at: rs4553604 - Blue.
CT at: rs989869 - Contrasting sphincter around pupil.
CT at: rs3947367 - Contrasting sphincter around pupil.
AA at: rs4778138 - Weak Amber Gradient
TT at: rs1129038 - Penetrance modifier. Blue.
GT at: rs1470608 - Medium melanin on Anterior Epithelium. Gives dark eyes.
GG at: rs16891982 - Starburst (Collarette)
TT at: rs1667394 - Starburst (Collarette)
CC at: rs12203592 - No pigmented Collarette.
CC at: rs4779685 - Flecks (Nevi).
AG at: rs11634406 - Flecks (Nevi).
GT at: rs1448485 - Inhibits weak amber gradient.

Volat
02-23-2018, 06:19 AM
I forgot to say 23andme is wrong. I have gray colour eyes. The colour of my eyes were blue in my childhood.


My eye colour is similar to this colour


http://cdn.playbuzz.com/cdn/3b101102-a8c0-4786-83d4-2875c27b2c0a/44a8b5bb-779b-4f1c-8a55-02c0d2831791.jpg

poi
02-23-2018, 07:41 AM
Yo poi could you run mine too

Will do, but I want to make sure my process is right because I'm getting 100% white (lol) using those 6 SNPs I have from ftdna and 23andme.

George Chandler
02-23-2018, 08:52 PM
Let me go ahead an make a Public Service Announcement PSA:

In the above calculator rs16891982 is backwards (notice it says reverse in the website). That means that if you are GG you should enter CC; notice that the allele frequency in the parent populations are backwards with C being solely found in 84% of White; and G being found in the majority of Intermediate and all(95%) of the Black. Go ahead a correct this to get appropriate results. If you are heterozygous then is ok; but be mindful of the allele frequencies.

Being that SLC45A2 can influence skin, hair and eye color isn't both the "CG" results for poi (or GC) depending on how you display it the mutation for hair pigment..or is SLC45A2 combined with another pigment defining pair of Alleles defining the starburst? The ancestral Allele is "C" but it's paired with another "benign allele".

George