Ancestry Testing: Sallis Right Once Again

Surprise!  Emphasis added.

First, Happy Fourth of July!  Enjoy your fireworks!

See this.

How is it possible that identical twins have different DNA results? It is not only surprising but shocking as well. If you do not live close to identical twins, it can be challenging to identify them. It follows that identical twins have the same DNA.

However, in a somewhat surprising incident, identical twins got different results from their DNA tests. If the tests were 100 percent accurate, then the chances of identical results were also high.

Obviously the tests are not “100 percent accurate,” regardless of what heavy-breathing fetishists believe when they sweatily post their results.

Since the Agro sisters are identical twins, one would also expect identical results from their DNA tests. That is not the case, and the results have some differences. According to the Living DNA results, Charlsie has DNA from Scotland while Carly’s ancestry is from England.

Charlie is 10 percent more “broadly European” than Charlsie according to 23andMe. On the other hand, MyHeritage gives the twins Balkan heritage 60.6 % & 60.7% while FamilyTreeDNA gives them Middle East DNA of 13 to 14 percent. AncestryDNA gives them Eastern European culture of 38 to 39 percent. There is no reason why results from these five DNA tests could not be identical if they were 100 percent accurate.

While statistical error can occur, the wildly different results should give pause to anyone with a triple digit IQ (most “movement” “activists” are therefore excluded).

According to Mark Gerstein, the twins should get identical results since they would be sending the same raw data to the company. 

Within statistical error.

Gerstein is a Yale University, a computational biologist…

Who doesn’t understand statistical error variation?

… whose research proved that raw data for the twins was (sic) statistically identical. In case of the differences in results from the five tests, there are different explanations.

Reasons For Differences In DNA

The first reason for the different results of these twins is probably because the DNA databases consist of different reference panels. These are DNA samples of previous clients, and they vary from company to company. The reference panels also vary across borders, and each company has a different reference panel.

Exactly – exactly what I have been saying for years. This is likely the major reason for the differences.

As a result, these companies are likely to provide different ancestry results to the same customer. It follows that the accuracy of DNA results mainly depends on a large number of people getting DNA tests from the same company. The other issue is that your results can also be accurate if other members from your lineage exist in the reference panel.

“Other members of your lineage” can mean co-ethnics. This is particularly true for ethnies that are relatively genetically homogeneous – like the many of the Herrenvolk (or the Ashkenazi). In those cases, customers are essentially being compared to themselves – and then when they get results close to, or at, 100% (of what they want), they breathlessly present the data on “movement” sites.  It’s only slightly more meaningful if they had been compared to their own DNA.

Therefore, the results for people who come from underrepresented groups in the reference panels are likely to be less accurate. 

OK, did you read that?  A mainstream analysis says the same thing as the “insane” “paranoid piece of crap” Sallis.

Before getting a DNA test, you must make an effort to investigate the size of its database. The larger it is, the higher the chances of getting accurate results.

OK.

Different Algorithms

The other issue is that a DNA sample consists of about three billion parts. However, ancestry DNA companies tend to focus on about 700, 000 parts only to identify the genetic differences. The companies then use an algorithm which compares these 700,000 DNA parts to those found in the reference panel.

The algorithm used tries to find the closest match from the reference panel. It means that the level of accuracy depends on the size and quality of data in the company’s reference panel.  The aspect of the diversity of data in the reference panel can also affect the accuracy of your results.

Exactly. Also note that some companies present their data by default at the “50% confidence level.” Flip a coin!  If you increase the confidence levels to, say, 90% (that is still lower then the 95% typically used in most scientific literature), you’ll then see the “unassigned” inflate to high levels for some people – they don’t have good reference matches.

If the reference panel consists of a sample of people who are predominantly from a particular region, accuracy might not be reliable. The reference panel should include data from diverse areas to increase the chances of the accuracy of the results. The sample should be more representative and cover large areas.

The solution is to increase the diversity of the reference panels. OK, if “Italian” samples are originally all from the North (plus Sardinia), you can add Southern samples. But that may make the reference too broad.  Why not then have reference samples for each region of Italy, and the same holds for different regions of Germany, France, the UK, etc.?

You should also not quickly take DNA results as conclusive since many companies assign different definitions for the region. Many DNA companies mainly cover areas consisting of European ancestry as reference panels.

Yes, but not all European ancestry regions are equally covered.

Some companies are however making inroads into other areas.

When they expand these “other areas” while not increasing their European coverage, that inflates “admixture” for those Europeans not well covered, for the reasons explained above.

You need to be careful about companies that claim to trace your ancestry to your home town or country. It may not be surprising that such company’s reference panel does not have many samples from your country. However, they will still try to find some close matches even though they are not accurate.

Indeed. A Bill Clinton or Anthony Weiner will be well represented in the databases by similar co-ethnics, a Codreanu or Evola not so much.  Note that “Italian” samples from population genetics databases are typically Bergamo, Tuscany, or Sardinia, as opposed to, say, Campania, Calabria, or Sicily. When sub-regions are included – say Northern Italian vs. Southern Italian, these are typically mixes of the narrower areas within the sub-regions, which doesn’t always work for more genetically heterogeneous Southern European populations. 

Problems exist (to a lesser extent) even for some Northwest Europeans. If Derbyshire is of English genealogical descent, why did he get only ~70% “British/Irish?”  Is “British/Irish” too broad a category for each subunit of that category?  However, given the greater genetic homogeneity of Northwest Europe, the remaining 30% of his ancestry was from closely related groups, so while sub-optimal, it’s not completely catastrophic. Others are not so lucky.

The other factor is that no governing board ensures the validity of DNA results. All companies keep the data to ensure privacy, so it is difficult to test its efficacy by independent players. The tests are not subject to a standard for testing or diagnosis like in the medical field.

OK.

Limitations of Ancestry DNA Tests

The ancestry DNA results are subject to change if there are additional reference samples. 

Exactly. Notice that movement” droolcups never question that accuracy of a given set of results even when the results are changing every time the company’s database is updated.

In most cases, these results are based on estimates; hence you should treat them with caution.  

In other words, don’t post them as if they are absolute proof of your “purity” (or lack thereof), excitedly talking about data points to the decimal point.

The other thing that you should note is that many reference groups consist of people self-reporting their ancestry. If you are a minority and do not belong to their group, you may get inaccurate results.

Exactly.  And “minority” can be interpreted to mean “minority” with respect to the reference databases.  

This all doesn't mean that the tests are completely useless, but they need to be evaluated in the proper context.  Unfortunately, "movement' retards seem incapable of doing this.