What is the accuracy of yDNA matches?

Question

The section of help under -  Does "Confirmed with DNA" mean a relationship is proven?

1. Even if a confirmation method is perfectly applied, there are limitations to DNA testing. For example, if a man is an exact yDNA match with the man who raised him, the paternity can be considered Confirmed with DNA. However, it's possible that the real father is the supposed father's brother, father, paternal uncle, etc., since they would all have the same Y chromosome.

The limitations described here are way too narrow.  It's possible the real father is any male who shares the same yDNA markers.  This would include everyone who matches on ysearch or FTDNA.

This example may confuse people in associating the guidelines used by the DNA services to help find how two people are connected. This requires knowledge of surnames.Guidelines involving Non-Paternal events assume no such knowledge.

If I have an exact yDNA match with an 8th cousin. This allows me to confirm with DNA, my father.  If an autosomal test between my father and me indicated no relationship, then the real father could be anyone who shared those same markers since we don't know the surname.

Comments

Another point I noticed.  If a Non-Paternal event occurred, then it is highly UNLIKELY that the real father is the supposed father's brother, father, paternal uncle, etc., even if they did all have the same Y chromosome. We don't know the yDNA values in a Non Paternal Event for someone not tested.

Answer 1

Can you show an actual example in WikiTree where a male is a Y-DNA match (in YSearch) with his distant direct paternal line cousin but he and his father's auDNA tests shows no relationship (in GEDmatch)?

There are hundreds of Y chromosome surname DNA projects which have demonstrated the value using Y haplotypes in genealogy.

Comments

Peter, I don't believe I ever suggested that yDNA is not valuable to genealogy.

The specific statement is describing the "limitations to DNA testing". The example seems to imply some close relationship based on surname.  The actual limitation of the DNA testing in a case of a Non Paternal event is the real father shares the same number of markers as another male who may or may not be closely related.  The probability increases with the same surname, but the example implies some closer relationship, which would minimally impact the existing tree.

I recently discovered a NPE in which I am actually a Frost. I had to use yDNA, and multiple auDNA tests to identify the single individual. I discovered my paternal line is Frost (great grandfather).  Years ago, a carpenter named Frost, and well known to the family, worked on my house over a period of months. I have since determined via a paper trail, he is my 4th cousin. 

Both our families stayed in the area for generations. 

Hypothetically, he could have been the biological father of my son. I would have been an exact match as in my example, but the relative could have realistically been much further apart. 

I am only suggesting a documentation change to better reflect the limitations and possibilities.

---

Great point.

Answer 2

Interesting choice of words... Accuracy is a technical term that is more likely best expressed in the example provided as "likelihood".

Accuracy is best used to describe two events - the read results of the DNA sample provided, and the interpretation of those primary results against a known set of markers (essentially positions on the chromosomes) including SNPs and STRs. Even these markers can have unreliable locations or shared locations that make confirming their presence less accurate technically. That is why multiple reads of the same chromosome position are used by testing companies to reduce the likelihood of a false positive result. A lower number of reads reduces the 'accuracy' of the result versus a higher number of reads at the same location. Even confirmed positive results with as few as 3 reads can carry a sizable error component of 20 to 30%! In this sense there high accuracy but less than perfect likelihood.

Test accuracy is built into each of the three major tests - yDNA, mtDNA, and atDNA. But how accurate are the results against the phylogenetic tree?

Well there are several different trees - YFull, ISOGG, etc. Where a testing company places your results in relation to these ever-changing and growing trees is another accuracy issue. Generally speaking you are not too far off from where you belong and the better companies will reassess your place with new info and move you to the right place over time. So I would say the label placed on your results is as accurate as it can possibly be - especially in relation to any other person who shares the same test results (which is to say there is no relative difference between you).

But there are magnitudes of difference in the accuracy of how different test results may be used to establish relatedness between or among different people sharing results. yDNA tests like next-generation sequencing offerings such as the BigY are the most accurate in that specific SNPs are identified up and down your individual family line that essentially pinpoint your distant ancestry as well as your recent genealogical timeframe. Testing for STRs - from 37 to 111 - have great test accuracy but more limited accuracy with regard to matching one's results to others. mtDNA full sequence tests provide the greatest level of detail among members of the same maternal line but matching is more difficult in that there is no great measure of when that maternal line was shared - the resolution of a specific generation is lost without great paper documentation for anyone comparing results. Lastly, atDNA can be fun and exploratory but far less accurate in it's application. The shorter the lengths of shared DNA snippets, the fewer the snippets, the lack of great paper documentation, etc all compound and reduce the potential accuracy of these matches. At its extreme it appears to be no better than 'reading tea leaves' and at its best one will share only 50% with a parent, sibling, or aunt/uncle. A similar dearth of likelihood may be said to exist when comparing 12 to 37 STR markers of a yDNA test.

Comments

Maybe a better term than "accuracy" could have been used.  :)

When the "Father is Confirmed with DNA", the guidelines state the Parent-child relationship has essentially confirmed the relationship using Traditional methods. This means that unless DNA exposes a NonParental Event,  The current tree is considered complete.

A DNA test can only further support the existing parent/child relationships, or refute it entirely.  Using the term "Confirm" implies conclusiveness, but it is not conclusive.

I am not sure how often a ystr-12 gentic distance 0 that matches along a completely filled confident tree connection will then expose a Non Paternal Event when upgrading to ySTR 37 match.  It will clearly expose an NPE if either don't match. Just thinking out loud...:)

---

Good points Ken!  I was not around in 2013 when the DNA guidelines were written at WikiTree but a lot has happened since then...

---

Ken, I've posted about accuracy of a 100% match for different number of markers in another thread just now. But here's the link again to FTDNA: https://www.familytreedna.com/faq-markers.aspx

---

Andreas, Thank you, you may also find these pages useful because it looks like they were done about the same time as the 111 str release.

ySTR 111 Comparisons
ySTR 67 Comparisons
ySTR 37 Comparisons
ySTR 12 Comparisons

The text contained on your page indicated the reason for introducing the 67 marker test was because some families could reach their objectives with less markers but that 67 markers would help more families.

I have a 111 marker comparison with a known ~9th cousin and quite a few about 6 Frosts at the 67 marker comparison, all but 1 is a 0 degree.

The genetic clock in my family either has not hit within these generations or that the clock caused the same mutations at the same time. 

For me, a 12 Str test would be just as good as the 111 test. IMO, the most important objective for the yDNA and mtDNA is the possible presence of an NPR.  I don't believe higher tests will make much difference in my case. The 37 marker test with 0 differences indicates a probable match within 8 generations.  For the purpose of wikitree, I don't see any measurable difference between the two tests other than cost.

When you already know the genetic distance and it only needs 90% of the markers to match, I think by recommending people start with the 12 marker test, would generate significantly more testing. 

I think a y-STR test costs about $69 and a 37 marker test about $169.  

The only reason I upgraded to the 111 marker test was to help find how I was connected to a couple of Frosts. It was not for confirming with DNA. This upgrade really didn't help much in my case but I would still recommend it to others in these cases.

Answer 3

Great example Ken, haven't thought about this but it's good that you're thinking out-of-the-box!

The proper solution should be to use an autosomal DNA test as well and use the shared DNA vs expected shared DNA as further evidence to come to the "confirmed by DNA".

Chris won't be happy but you're rightfully pointing out a flaw in the current rules for "confirmed with DNA" on Y-DNA tests!

Comments

I would suggest another solution by using "further supported with DNA" to better allow for differences in accuracy that comes with the different DNA tests as they apply to the different accuracies that are associated with the different degrees of separation.

This would allow the yDNA and mtDNA tests to be used in ways restricted by "Confirmed with DNA".

I think this will make the point.

You should be able to track partial process. A match wth a known relative, using predictions that support the documented relationships, up to by not including the MRCA's would be tracked.  If the exact relationship between these two matches is the objective, then we could accomplish this in multiple ways.

The auDNA only solution would require two additional tests. One match that is more distant than the male MRCA and one match that is more distant than the female MRCA.  "Confirmed with DNA" allows for other matches, and only requires one additional match.

There is also a belief that by looking at the segments, a different and better conclusion might be reached. IMO, the existing rules need some explanation.

Here are some other differences. A 3rd cousin 1x moved would most likely not support a match with only a single 7cm segment. There is no requirement that the additional tests have a triangulated segment.

We should be able to use partial auDNA to get part way, and then use yDNA and/or mtDNA to complete the process.

Instead of using the audDNA only solution, we should be able to use yDNA and mtDNA matches with the partial auDNA.

In other words, (1) the parent/child relationship for those profiles up to but not including the MRCA using auDNA, and the last two profiles using mtDNA and/or yDNA.  

The current restrictions do not allow for this.

---

I just had a lightbulb moment....

I was wrong in thinking the only reason the requirement to use segments, especially triangulated segments was to locate a common ancestor by someone matched but who does not know how they are related.

Why only require 1 segment that has to be triangulated, and why doesn't the genetic distance predicted by the DNA Services play any role?

From a match of a third cousin 1x removed, we can say that we have support to show AT LEAST a 1/2 3rd cousin 1x removed. The only difference using the current guidelines is that you can identify the common ancestor but you can still only state that you can show AT LEAST a 1/2 3rd cousin 1x removed.

Why not require 2 triangulated segments to prove the matches exact relationship?

The current rules are used to  emulate ySTR and mtSTR matches using auDNA and then utilize them the same way as ySTR.  These rules proved a 1 to 1 connection the same way ySTR operates.

This explains why those auDNA segments that determine genetic distance are ignored.  The one segment length is just about as accurate as a yDNA match.

This is why you need to look at the segments. You can't emulate ySTR without it.