Problems with commercial DNA tests.

[Slavic Italian]

Notice some are happy with their updated 23&me results while others are not? It's because certain population panels are not well represented. Also, certain SNPs are left out of the testing as well.


Most commercial DNA tests rely on reference panels made up of individuals with known ancestry, but those panels are often unevenly distributed across the globe. That means some populations are richly represented, while others are barely sampled or entirely missing.

[Scandal]

Gedmatch (and Vahaduo) oracle result > Commercial dna test result, that was always the case. You've been a member since 2016, so you probably know that. They just involve too much DIY which make them unpopular.

[JerryS.]

I tend to get better results with modern individuals using my G25 than I do with anything else.

[calxpal]

Yes I realize that the results will vary across the board for all sorts of reasons. My biggest issues with the new 23andme update though is IMO how inconsistent the assignments are, especially when you compare with family members and look at the parental phasing which used to be one of their best features imo and now it is useless. And how the "most likely" setting seems like a ridiculous crapshoot to appeal to a small minority of customers, forcing ancestry to be assigned to categories where it really shouldn't belong and having even slightly more distant parts of your ancestry gobbled up by one main category which is an issue that is more common to 3rd party DNA sites. The ancestry timeline is absolute nonsense now too (for me). When pre-update my results were almost perfect and the information and interface was good.

I think 23andme rushed this update and used AI 100%, also how are they less confident with my European assignments now than ever before (60% Broadly European at 90% confidence!!!!!). Which sucks because with all the new subregions I thought it had a lot of potential but seems to often oversimplify for people with varied ancestry now. Not to mention I've noticed multiple bugs and glitches in the chromosome painting and the ancestry composition overall. Also the new colors are hard to read. Also me and my sister both have random ancestry assignments that neither of our parents have, in a reasonable amount.

23andme has gone the wrong direction with this update, or they needed way more time to work on it.

I do think over and under sampling may indeed have been an issue with this update as well.

The way this "most likely" setting tries to force chromosomes to be nearly one color though just feels unscientific though. I will say the results for me at 50%-60% confidence are not as good as v6.0, but are much better than the "most likely" results in terms of matching my background which makes me wonder why they did not just keep the 50% default which seems to work better and instead use this highly experimental new setting.

I also lost a few genetic groups too so less information as well. I was hoping they would improve on these or at least add a new feature in the update, especially considering how long it's been. I'm not gonna be that loyal to 23andme anymore as this update worries me about the future of it.

It seems like getting good detailed results with this update is based on luck.

People can disagree with me though and say that "oh I just don't like my results" or that I'm being too picky and having high standards...but I realize my opinion is subjective still and have seen a lot of good results out of the update. Or people will try to tell me my genealogy is wrong....well my results made more sense up until now, and so for a few of my family members with the update too.

I do think they did indeed actually do a good job with the Eastern Europe and Balkan categories relative to others.

Also as a note I have been a 23andme customer for almost 9 years and have v4 and v5 and have seen many updates, and this one is by far the most off-color and poorly refined IMO. I can only hope they try fix it up but maybe not for years.

I man sometimes it is fun to see things different from what you know in the results, and it can make sense, but it also has to be realistic and consistent and not misleading or seemingly random.