https://youtu.be/rW_O0iRUtr4
Printable View
I am very skeptical regarding your phenotype predictions after it predicted olive/dark skin and some sort of Armenoid phenotype for me.
are you also skeptical of hirisplex? i didnt rely on my phenotype predictor at all in the video, only hirisplex. Also the phenotype predictor is separate from NOSHACOT. my NOSHACOT predictions are the best out of every tool that exists.Quote:
Originally Posted by Vessna
Also if you scored olive skin with my tool, you'd also score it with other tools that exist, such as hirisplex. have you tried running yourself through it?
I just found your report. Had to search through my email sent folder pretty far back to find it. You didn't score Armenoid at all. Your closest 2 way was Norid + pamirid which is very possible for a belarusian.Quote:
Originally Posted by Vessna
You also didnt score dark skin. Your result was 96% olive for skin tone.
I entered in some of your genotypes from the NOSHACOT table on your trait predictor report into the online hirisplex prediction website. Since I didnt store your DNA file anywhere, I had to manually enter some of your genotypes into hirisplex.
This is what you score with hirisplex:
p-value AUC Loss
blue eye 0.12 0.002
intermediate eye 0.143 0.006
brown eye 0.738 0.002
blond hair 0.405 0.015
brown hair 0.451 0.014
red hair 0.007 0.052
black hair 0.137 0.003
light hair 0.755 0.001
dark hair 0.245 0.001
very pale skin 0.008 0.03
pale skin 0.266 0.022
intermediate skin 0.708 0.04
dark skin 0.015 0.008
dark to black skin 0.003 0.005
Now let's compare that with your trait predictor result:
Eye color likelyhood distribution based on whole genome data:
Blue eyes : 0.233%
Blue eyes with Amber Center Ring : 22.125%
Green eyes : 12.620%
Hazel Eyes : 48.973%
Brown Eyes : 15.805%
Darkest Brown Eyes : 0.244%
Hair color likelyhood distribution based on whole genome data:
Red hair : 0.049%
Light Blond hair : 0.007%
Dark Blond hair : 0.027%
Light Brown hair : 5.373%
Dark Brown hair : 93.775%
Black hair : 0.768%
Skin color likelyhood distribution based on whole genome data:
Palest Skin : 0.283%
White Skin : 1.249%
Olive or Mediterranean Skin : 96.648%
Light Brown Skin : 1.819%
Dark Brown Skin : 0.000%
Hair Texture:
Straight hair : 35.177%
Wavy hair : 43.057%
Curly hair : 21.620%
Kinky hair : 0.146%
Nose Shape:
'Greek' nose shape likelyhood : 75.315 %
'Snub' nose shape likelyhood : 24.685 %
Seems like they match pretty well!
In fact your result with trait predictor is pretty much exactly the same as with hirisplex.
And it would be the same with any other good tool you would use.
Match pretty well with what? I don’t have dark brown hair nor I have olive/Mediterranean skin. My eyes are green hazel, that’s probably the only correct thing in this report. 23andMe actually predicted everything correctly: hazel eyes, fair/light skin, dark blonde/light brown hair.Quote:
Originally Posted by AndreiDNA
You also predicted blonde hair and white/light skin for my husband where in reality he has intermediate skin that tans unlike mine that burns, and he has medium to dark brown hair.
My “Mediterranean” skin :D
https://i.ibb.co/qMpX8SQP/05225-CB7-...BE2925-EC2.jpg
Let's do a side by sideQuote:
Originally Posted by Vessna
Skin color: my tool predicted olive, hirisplex predicted intermediate (also olive)
Hair color: my tool predicted dark brown, hirisplex predicted brown (hirisplex is closer here)
Eye color: my tool predicted hazel, hirisplex predicted brown (my tool is closer)
For different people, different tools will work better. My tool is better for most. Sorry you didn't get what you were looking for, but don't try to discredit what I created by lying about your results!
"You also predicted blonde hair and white/light skin for my husband where in reality he has intermediate skin that tans unlike mine that burns, and he has medium to dark brown hair."
the prediction for your husband was split between dark blond and light brown. Saying it was just "blond" is a bit incorrect.
Also, your husband scores this with hirisplex:
p-value AUC Loss
blue eye 0.848 0
intermediate eye 0.088 0
brown eye 0.065 0
blond hair 0.704 0.017
brown hair 0.25 0.019
red hair 0.019 0.071
black hair 0.028 0.007
light hair 0.959 0.005
dark hair 0.041 0.005
very pale skin 0.021 0.036
pale skin 0.469 0.02
intermediate skin 0.506 0.04
dark skin 0.003 0.006
dark to black skin 0 0.004
Let's compare your husbands results with hirisplex vs my tool now.
Eyes: hirisplex: blue, my tool: green
Hair: hirisplex: light blond, my tool: dark blond to light brown
Skin: hirisplex: intermediate to pale, my tool: White.
Once again, my tool isn't any worse than hirisplex for your husband.
But thats just you and your husband. For myself, noshacot is much better than hirisplex.
I score blond hair, just like your husband, with hirisplex, but light brown with my tool
I score blue eyes with hirisplex, but green/blue with amber center with my tool
Guess what, my hair is light brown and my eyes are green.
Statistically speaking my tool is better than hirisplex in most cases. Once again, sorry you didn't like your result, but my tool isn't any worse than its competitors.
I’m not sure why you’re accusing me of lying. My 23andMe report literally states that my genetics make me “most likely to have lighter skin.” It assigns a combined 71% probability to very fair and moderately fair skin, 25% to light beige skin, and only 3% to olive skin.Quote:
Originally Posted by AndreiDNA
That’s why I said 23andMe was more accurate in my case. I’m fair-skinned and burn easily. We can disagree about which model performs better overall, but my interpretation of my own report is consistent with the results shown.
https://i.ibb.co/HLYnDHmd/IMG-5958.jpg
My predicted phenotype with yseq pheno predictor:
Attachment 148656
Vs my Trait Predictor:
Attachment 148657
Vs my actual phenotype:
Attachment 148658
My result with hirisplex:
PBlueEye 0.9110910019
PIntermediateEye 0.05659703369
PBrownEye 0.03231196441
PBlondHair 0.5651961134
PBrownHair 0.325021255
PRedHair 0.05878985349
PBlackHair 0.05099277807
AUC_Loss_HairShade 0.006540924711
PVeryPaleSkin 0.02444099902
PPaleSkin 0.4459833968
PIntermediateSkin 0.5221414264
PDarkSkin 0.006253594033
PDarktoBlackSkin 0.001180583755
Vs my trait predictor noshacot result:
Eye color likelihood distribution based on whole genome data:
Blue eyes : 26.403%
Blue eyes with Amber Center Ring : 55.121%
Green eyes : 13.560%
Hazel Eyes : 4.812%
Brown Eyes : 0.105%
Darkest Brown Eyes : 0.000%
Hair color likelihood distribution based on whole genome data:
Red hair : 0.538%
Light Blond hair : 4.724%
Dark Blond hair : 27.083%
Light Brown hair : 54.808%
Dark Brown hair : 12.633%
Black hair : 0.215%
Skin color likelihood distribution based on whole genome data:
Palest Skin : 13.478%
White Skin : 70.942%
Olive or Mediterranean Skin : 15.576%
Light Brown Skin : 0.003%
Dark Brown Skin : 0.000%
My eye color, in case you cant see it in the face shot:
Attachment 148659
Since we’re talking about phenotypes, can you explain why my closest pheno match was some type of Caucasus/Armenoid? I’m genuinely curious about how the tool arrived at that result.
I can’t post my full face for privacy reasons, but here’s my eye and hair color.
https://i.ibb.co/LDcCLL9T/IMG-1850.jpg
https://i.ibb.co/84xj1pWn/IMG-1990.jpg
https://i.ibb.co/HD2fKMjF/IMG-1183.jpg
https://i.ibb.co/mVQCCQZQ/IMG-0725.jpg
the way my phenotype predictor works, mtedid (your #1 pheno) is nowhere near armenoid. It is a intermediate pigmentation caucasoid type with long nose, much more similar to carpatid than armenoid.Quote:
Originally Posted by Vessna
You scored hazel eyes (that aligns with mtebid), dark brown hair (once again aligns with mtebid, olive skin(once again aligns with mtebid, not with armenoid though), wavy hair (aligns with mtebid), and a strongly greek nose shape (once again aligns with mtebid).
My phenotype predictor uses the qualities listed above ^ together with morphology-related genotypes to predict phenotype.
For morphology calculator, you scored 5.5% NE Europe, 60.2% Mideast/West Asia, 6.63% West Europe, and 27% South Asia.
When you combine the results of these two calculators you get a light west asian pheno, such as Mtebid, which is what you scored.
You keep saying "Armenoid" though... When your result is in fact very very far from it.
That’s part of why I’m skeptical of the tool. It seems to be assigning traits that don’t actually match me. I don’t have dark brown hair, my eyes are green-hazel rather than hazel, and my nose is not “strongly Greek”, it’s not Greek at all.Quote:
Originally Posted by AndreiDNA
The morphology calculator result is even harder to reconcile with my actual ancestry. I am almost entirely northeaster European, not 60% Middle East/West Asia and 27% South Asia.
If the calculator is producing results that contradict both my observed traits and my DNA results, why should I assume the phenotype prediction is accurate rather than the model simply being off?
Statistically, a big portion of europeans will score the way you do on the morphology calculator, simply because allele frequency wise, middle easterners and south asians are similar to europeans (in the global aspect, compared to say sub saharans or east asians).Quote:
Originally Posted by Vessna
"I don’t have dark brown hair, my eyes are green-hazel rather than hazel, and my nose is not “strongly Greek”, it’s not Greek at all"
Hair color was off, but not by much compared to hirisplex.
Your eye color was actually spot on.
Blue eyes : 0.233%
Blue eyes with Amber Center Ring : 22.125%
Green eyes : 12.620%
Hazel Eyes : 48.973%
Brown Eyes : 15.805%
Darkest Brown Eyes : 0.244%
^this exactly aligns with green-hazel eyes.
Regarding the nose shape, you might be misinterpreting your nose shape or it might be a case of the tool getting your trait wrong. It got the nose shapes in my family 100% right. Including the people I know irl whose genomes I have.
You can assume anything and it doesn't much bother me. I see you've already assumed all you could and even lied openly about your results, claiming to have received an "Armenoid" result when in fact your predicted phenotype (top 2 way model morph) looks like this:
Attachment 148660
Attachment 148661
I never looked at your tool until this thread. It is way off for Vessna so I wouldn't trust it at all. Vessna is one of the fairest women on the forum, along with Kestrel and nittionia.Quote:
Originally Posted by AndreiDNA
well its good for most people who have used it. I wouldn't have sold the executable over 300 times (before eventually stopping sales) if it wasn't good.Quote:
Originally Posted by ~Elizabeth~
I didn’t lie about anything. I described the result as I understood it from the categories and explanations provided by the tool.Quote:
Originally Posted by AndreiDNA
I’m happy to admit that the eye-color prediction was fairly close. My issue has never been that every single prediction was wrong. My issue is that several of the key traits don’t match me, including the hair color, skin tone, and the nose description.
Saying that I might be misinterpreting my own nose isn’t really evidence that the prediction is correct. At that point we’re just assuming the calculator must be right and my observation must be wrong.
As for the morphology calculator, I understand that allele frequencies can overlap between populations. My point is simply that the output doesn’t appear to reflect either my known ancestry or my actual appearance particularly well, which is why I’m skeptical of its accuracy in my case.
I think people can be obsessive about DNA and are desperate or curious to try every tool they can find. I tested with 23andMe twice, DNATribes countless times (always buying the updates), MyHeritage, AncestryDNA, FTDNA, exploreyourdna, Illustrative DNA, another one I forgot the name of, and GEDmatch (all the calculators).Quote:
Originally Posted by AndreiDNA
By the way, here is exactly what your tool predicted for me
https://i.ibb.co/XnTzW8b/IMG-0147.jpg
I saw a photo of you once in the Member's section that was soon deleted. What this tool predicted for you is way, way off. Looks nothing like you.Quote:
Originally Posted by Vessna
Hair color is not 100% genetic. Large-scale twin studies estimate that the heritability of natural hair color falls somewhere between 73% and 99% depending on the specific color (for instance, black and red hair are more rigidly genetic, while varying shades of brown and blonde are more susceptible to other factors). The remaining variance is driven by environmental factors and time.
this isn't armenoid. And this here was the closest 2 way morph, which looks even more european:Quote:
Originally Posted by Vessna
Attachment 148664
Whether the morph is classified as Mtebid, Carpathid, Armenoid or something else isn’t really the issue for me. The issue is that I don’t think the prediction looks much like the person it was supposed to be predicting. That raises uncomfortable questions about the accuracy of the calculator especially when these predictions are used in YouTube videos to make claims about ancient phenotypes. If the model cant accurately predict the appearance of a living person whose phenotype and ancestry are known then it’s reasonable to be skeptical about its ancient reconstructions.Quote:
Originally Posted by AndreiDNA
Edit: I don’t look like that updated morph either, not even close.
5.5% NE Europe, 60.2% Middle East/West Asia and 27% South Asia for a person who is overwhelmingly Northeastern European? I’ve taken multiple genetic tests and this result is just nonsense.Quote:
Originally Posted by AndreiDNA
If a calculator can turn a Belarusian into 60% Middle Eastern and 27% South Asian then I think skepticism is more than justified.
Could it be that your sample and your husband's sample were accidentally switched?
When it comes to the phenotype predictions, they may not be fully refined, but the recombination patterns responsible for pigmentation are fairly well understood. You can check a few of them yourself in the Excel file.
this isn't an ethnicity calculator. this is a morphology calculator. it's not supposed to yield 100% NE europe for every belarusian, its supposed to yield whatever your morphology resembles.Quote:
Originally Posted by Vessna
According to Vessna, these two belarusians should score the same thing (100% NE Europe) for morphology
Attachment 148666
Attachment 148667
Quote:
Originally Posted by cass
No, the samples definitely weren’t switched. Those are my results and the tool came up with a questionable phenotype prediction for my husband as well.
Even if pigmentation genetics are relatively well understood the tool still failed to predict my pigmentation. It gave me a greater than 90% probability of olive skin which doesn’t match my actual appearance and also doesn’t match my 23andMe results.
My bigger concern is how these probabilities are being used. The calculator gives a range of possible traits, each with its own probability. A person can end up anywhere within that range, including outcomes that aren’t the most likely. Taking only the highest-probability traits and stitching them together into a single reconstructed phenotype can easily result in a face that looks completely different from the real person. This is what I should look like according to AndreiDNA:
https://i.ibb.co/ZPKgZfV/IMG-5961.jpg
I wouldn’t mind looking like this but I look like an ordinary potato. There are many examples online where even 23andMe get trait predictions wrong or only partially right despite being based on well-studied genetics. Predicting someone’s appearance from DNA is very difficult.
The problem is even bigger when the same approach is used to reconstruct ancient samples. In my case the tool classified me as Mtebid which contradicts my actual appearance and my actual ancestry. It also produced a questionable result for my husband. If our samples were from an ancient Corded Ware burial those reconstructions could then be presented in videos as approximations of what CW individuals supposedly looked like :D
AndreiDNA predictions of phenotypes for me and my husband:
https://i.ibb.co/q3N8zqGz/IMG-5963.jpg
https://i.ibb.co/60My2Wy9/IMG-5973.jpg
That’s where my skepticism comes from. If his methodology has trouble predicting the appearance of living people whose looks are known I think he should be careful about how much confidence he places in reconstructions of people who lived thousands of years ago.
Let’s not be abstract here, I am asking you to explain a real result on a person with known genotype and ancestry.Quote:
Originally Posted by AndreiDNA
It was the intention of my tool for *certain Belarusians, who have a mediterranean-west asian morphology, like Lukashenko for example, to score 60% MENA on the morphology calculator. Not every single Belarusian is 100% NE European morphologically. There are belarusians with more southern phenotypes (coloring aside) too.Quote:
Originally Posted by Vessna
The morphology calculator is about 50-60 snps that are all located in morphology related genes, such as EDAR and PAX3. It is possible for a Belarusian to score 60% MENA on it because there ARE belarusians whose morphology is 60% MENA.
There is actually a very valid reason for why your husband's result is the way it is, and if you ever explored the report you'd see it.
"AG genotype in EDAR's rs3827760, Likely has shovel shaped incissors and either partial or full East Asian Ancestry
GG in rs6542787, no European EDAR alleles. Likely not a European
AC genotype in EDAR's rs260690, Likely not European facial traits. Rare for Europeans, common for East Asians.
CT in rs892540, slightly shorter midface length
GG in PAX3s rs7559271, decreased odds of protruding nasal bridge, more upturned nose
CC in rs4648379, larger nose size.
CC in rs938036, likely has mandibular third molar, lower odds of tooth agenesis (missing teeth)
AA in rs7702108, slightly thicker eyebrows"
From his facial morphology panel section.
Attachment 148668
Attachment 148669
Attachment 148670
Attachment 148671
Attachment 148672
Me with siblings and our predictions. Old pictures from 2021, theyre a lot bigger now
Quote:
Originally Posted by AndreiDNA
I still don’t think my original question has been answered.
I asked how someone who is overwhelmingly Northeastern European ends up scoring 60.2% MENA and 27% South Asian on your morphology calculator. Instead, you shifted to comparisons to Lukashenko whose results we haven’t even seen.
As for my husband, he doesn’t have shovel shaped incisors, and he certainly doesn’t look East Asian like the calculator produced. Listing individual variants isn’t the same as showing that the final prediction is actually accurate.
The same goes for the Lukashenko comparison. Without seeing his results it’s hard to know what relevance that example is supposed to have.
What I’m trying to understand is pretty simple: how does the calculator arrive at 60.2% MENA and 27% South Asian for someone whose ancestry comes back overwhelmingly Northeastern European across multiple DNA tests? What do those percentages actually represent and what evidence is there that this interpretation is accurate?
That’s the part I keep asking about.
Nothing wrong with your calculator. Its fine. She is just unhappy she got Mtebid or Pamirid + Noric, something that's probably not so standard (but exists) in the country she's from.Quote:
Originally Posted by AndreiDNA
This is called anecdotal evidence. I can DM you pictures of my entire family and it doesn’t look much different from yours.Quote:
Originally Posted by AndreiDNA
statistics. As ive said before, a portion of belarusians will score what you scored. As i've also said before, this is the intention of the calculator. To predict morphology, rather than just go based off ethnicity.Quote:
Originally Posted by Vessna
"As for my husband, he doesn’t have shovel shaped incisors, and he certainly doesn’t look East Asian like the calculator produced." God almighty Ural/Volgid is now east asian! So mtebid is Armenoid, Uralid is east asian, what's next! I'm tired of talking with you. Pointless conversation
I am unhappy because it’s inaccurate. Instead of admitting that the prediction tool is not correct I am being gaslighted that I look different. Are you people really this autistic?Quote:
Originally Posted by Live The Magic
Only that he’s not Uralid. Your prediction was incorrect.Quote:
Originally Posted by AndreiDNA
That's because the prediction tool is correct, in fact the best of its kind, and you're going in circles, offended the calculator didn't yield you a Baltic result. You didn't like your result. Most people do. Time to move onQuote:
Originally Posted by Vessna