Mutations and probabilty questions

[capeflier]

I am curious as to what point one can say with 100% certainty that two people who we thought were related are in fact NOT related. I have a case of results that progress as follows 12/12 22/25 29/37 54/67. The 2 individuals share the unusual 385b=11 value, that alone puts them in the under 2% of the population range. If the two individuals were within 2 generations do these results 100% guarantee a non paternal event or is there still a possibility no matter how remote that there is a relationship.

The core values have been verified by testing others in the family at different labs. We have family photos that show several family members that look just like several members of the other family. Needless to say these results were not expected. If these two individuals were actually not related why wouldn't the results be much clearer the other way. The typical random 67 marker test has mismatches well into the 20+ range. Also due to the unusual 385b=11 value even the 12 marker haplotype is unusual. 8 matches out of 139,000 at FTDNA 1 match out of 35,000 YHRD,

If there can be 2 mismatches in a 12 marker test and have two people actually related false negative could there be a 4 in 24 and 8 in 48 12 in 64 mismatch among two people who are in fact related. Are there any circumstances that could cause a large number of mutations in a single transmission Ie alcohol or illness.

Forensic Y-DNA testing for paternity for years has used lower resolution testing MHT 9 ,SWGDAM 11. I am curious if perhaps there are some unknowns about probabilities that preclude the use of these high 37/67 marker tests to exclude a person as related.
In short while these tests are usually used to find an unknown relative at what point can they be used to say that a known relative isnt a relative at all.

I am not so much trying to make a square peg fit in a round hole as trying to understand at what point we become certain that two people are not related because needless to say the consequences of being wrong are dramatic. Thank you very much for any input.
Mark

[Clochaire]

I am curious as to what point one can say with 100% certainty that two people who we thought were related are in fact NOT related. I have a case of results that progress as follows 12/12 22/25 29/37 54/67. The 2 individuals share the unusual 385b=11 value, that alone puts them in the under 2% of the population range. If the two individuals were within 2 generations do these results 100% guarantee a non paternal event or is there still a possibility no matter how remote that there is a relationship.

The core values have been verified by testing others in the family at different labs. We have family photos that show several family members that look just like several members of the other family. Needless to say these results were not expected. If these two individuals were actually not related why wouldn't the results be much clearer the other way. The typical random 67 marker test has mismatches well into the 20+ range. Also due to the unusual 385b=11 value even the 12 marker haplotype is unusual. 8 matches out of 139,000 at FTDNA 1 match out of 35,000 YHRD,

If there can be 2 mismatches in a 12 marker test and have two people actually related false negative could there be a 4 in 24 and 8 in 48 12 in 64 mismatch among two people who are in fact related. Are there any circumstances that could cause a large number of mutations in a single transmission Ie alcohol or illness.

Forensic Y-DNA testing for paternity for years has used lower resolution testing MHT 9 ,SWGDAM 11. I am curious if perhaps there are some unknowns about probabilities that preclude the use of these high 37/67 marker tests to exclude a person as related.
In short while these tests are usually used to find an unknown relative at what point can they be used to say that a known relative isnt a relative at all.

I am not so much trying to make a square peg fit in a round hole as trying to understand at what point we become certain that two people are not related because needless to say the consequences of being wrong are dramatic. Thank you very much for any input.
Mark

Mark,

We are talking about YSTR tests, right? Those can only address relationships in the direct male line--which is, of course only the tiniest fraction of genetic heritage. I mean, there's your Da's mother's father's mother, etc., etc...

For YSTR tests, typical rule of thumb would say GD of 20+ definitely rules out direct male line relationship in a "genealogical" timeframe. If you really are satisfied, as you seem to be, that there are no lab errors.

Conventional wisdom is that one should focus on overall similarity of the extended haplotype, rather than individual markers alone, even when unusual. If one focuses too much on individual markers, there is a chance of missing the combined statistical importance of many more markers in the overall pattern.

You should also consider whether any contradictory evidence may be available from SNP tests. The conventional wisdom is that even for idividuals with similar haplotypes, if SNP tests show incompatible haplogroups, there is no direct male relationship within perhaps many 1,000s of yrs.

More speculatively on my part, it was recently pointed out to me that there are some quite evident limitations on accuracy of popular, and I suspect linear-modeled, TMRCA calculators for deep ancestry.

Many smart people are tinkering around to develop better calculators, but the phenom is complicated, and science may not yet have comprehensive understanding of variables affecting population over many 100s of years.

Speaking only for myself, and based upon my first very naieve steps toward an exponential model calculator, I think it may still be possible that you share a male line ancestor in the last 1,000 years. Many people would strongly deny this idea, so take it for what it's worth.

This is probably not the timeframe you are considering, and you probably wanted a stronger level of assurance. But I think about situations like the one you presented here, from time to time, and these are my thoughts.

Jack

[Clochaire]

I am curious as to what point one can say with 100% certainty that two people who we thought were related are in fact NOT related. I have a case of results that progress as follows 12/12 22/25 29/37 54/67. The 2 individuals share the unusual 385b=11 value, that alone puts them in the under 2% of the population range. If the two individuals were within 2 generations do these results 100% guarantee a non paternal event or is there still a possibility no matter how remote that there is a relationship.

The core values have been verified by testing others in the family at different labs. We have family photos that show several family members that look just like several members of the other family. Needless to say these results were not expected. If these two individuals were actually not related why wouldn't the results be much clearer the other way. The typical random 67 marker test has mismatches well into the 20+ range. Also due to the unusual 385b=11 value even the 12 marker haplotype is unusual. 8 matches out of 139,000 at FTDNA 1 match out of 35,000 YHRD,

If there can be 2 mismatches in a 12 marker test and have two people actually related false negative could there be a 4 in 24 and 8 in 48 12 in 64 mismatch among two people who are in fact related. Are there any circumstances that could cause a large number of mutations in a single transmission Ie alcohol or illness.

Forensic Y-DNA testing for paternity for years has used lower resolution testing MHT 9 ,SWGDAM 11. I am curious if perhaps there are some unknowns about probabilities that preclude the use of these high 37/67 marker tests to exclude a person as related.
In short while these tests are usually used to find an unknown relative at what point can they be used to say that a known relative isnt a relative at all.

I am not so much trying to make a square peg fit in a round hole as trying to understand at what point we become certain that two people are not related because needless to say the consequences of being wrong are dramatic. Thank you very much for any input.
Mark

Mark,

We are talking about YSTR tests, right? Those can only address relationships in the direct male line--which is, of course only the tiniest fraction of genetic heritage. I mean, there's your Da's mother's father's mother, etc., etc...

For YSTR tests, typical rule of thumb would say GD of 20+ definitely rules out direct male line relationship in a "genealogical" timeframe. If you really are satisfied, as you seem to be, that there are no lab errors or so-called "nomenclature" issues in various lab reporting protocols.

Your reference to multiple lab tests for "core values" suggests to me that perhaps your mismatches relate primarily to "non-core" values? If so, I would take up the discrepancies with the labs. A fellow, I think named Calkins, wrote to this board with what may be a similar problem: Great match on 1st through 37th markers to known cousins, and fell to crap 38th through 67th. Turned out one lab just plain screwed up; no explanation as to why, but after much effort, a correction and apology were issued.

Conventional wisdom is that one should focus on overall similarity of the extended haplotype, rather than individual markers alone, even when unusual. If one focuses too much on individual markers, there is a chance of missing the combined statistical importance of many more markers in the overall pattern.

I am unaware of any phenom which would cause an "accelerated" rate of mutation.

You should also consider whether any contradictory evidence may be available from SNP tests. The conventional wisdom is that even for idividuals with similar haplotypes, if SNP tests show incompatible haplogroups, there is no direct male relationship within perhaps many 1,000s of yrs.

As an aside, and more speculatively on my part, it was recently pointed out to me that there are some quite evident limitations on accuracy of popular, and I suspect linear-modeled, TMRCA calculators for deep ancestry.

Many smart people are tinkering around to develop better calculators, but the phenom is complicated, and science may not yet have comprehensive understanding of variables affecting population over many 100s of years.

Speaking only for myself, and based upon my first very naieve steps toward an exponential model calculator, I think it may still be possible that you share a male line ancestor in the last 1,000 years. Most people would strongly deny this idea, so take it for what it's worth.

This is probably not the timeframe you are considering, and you probably wanted a stronger level of assurance. But I think about situations like the one you presented here, from time to time, and these are my thoughts.

Jack

[capeflier]

More information

Jack Thank you for your thoughtful reply.
Yes this is Y-STR. I am convinced because of the multiple tests with different labs that the actual values are correct. The other lab was 35/43 on a subject only 2 generations away again with that 385b=11 in common.
My father is deceased and the two familes do not share a common mother and they are both deceased..I do have some female half siblings but they have not been willing to take a siblings test. Also half sibling tests are notorious for being inconclusive,

If my Y-DNA results 54/67 35/43 could be considered inconclusive as opposed to positively negative I might be able to convince my siblings to test further. The labs seem so cut and dry that these results are on the not related side of their cut off points yet there seems to be evidence that there have been some exceptions. The evidence between the family photos and my mothers statements about my father are so compelling yet the science of DNA is leaving me mystified. I have seen evidence of paper trails confirming relations with 10/12 15/17 21/25 but no studies on high mutation levels on higher test levels. Im not inclined to just walk away without someone who knows more then I do giving me some input.

Thank you again for your response.

[Clochaire]

Jack Thank you for your thoughtful reply.
Yes this is Y-STR. I am convinced because of the multiple tests with different labs that the actual values are correct. The other lab was 35/43 on a subject only 2 generations away again with that 385b=11 in common.
My father is deceased and the two familes do not share a common mother and they are both deceased..I do have some female half siblings but they have not been willing to take a siblings test. Also half sibling tests are notorious for being inconclusive,

If my Y-DNA results 54/67 35/43 could be considered inconclusive as opposed to positively negative I might be able to convince my siblings to test further. The labs seem so cut and dry that these results are on the not related side of their cut off points yet there seems to be evidence that there have been some exceptions. The evidence between the family photos and my mothers statements about my father are so compelling yet the science of DNA is leaving me mystified. I have seen evidence of paper trails confirming relations with 10/12 15/17 21/25 but no studies on high mutation levels on higher test levels. Im not inclined to just walk away without someone who knows more then I do giving me some input.

Thank you again for your response.

Mark,

My first thought is that physical similarities could be driven by non-Y relationship.

My second thought is that the pattern of discrepencies you observe are reflective of "sampling error". Although we would like to place a lot of trust in the labs, and assume that if we cross-reference a handful of loci tested at multiple labs, the others should be fine as well. I took this approach once, but learned that it is not always so.

Long story short, I cross-referenced results from 3 popular commercial labs vs certified results from a state licensed lab. I found out that each commercial lab had 10 to 15 pct error in my case--and that all 3 got 1 of the same loci wrong.

This is probably why commercial labs won't even discuss their error rates, much less certify results.

So my question in this regard is;"When you look at the specific loci where you differ from this other fellow, are those cross-referenced multi-lab tested loci or no?"

Best of luck.

Jack

[efgen]

I am curious as to what point one can say with 100% certainty that two people who we thought were related are in fact NOT related. I have a case of results that progress as follows 12/12 22/25 29/37 54/67. The 2 individuals share the unusual 385b=11 value, that alone puts them in the under 2% of the population range. If the two individuals were within 2 generations do these results 100% guarantee a non paternal event or is there still a possibility no matter how remote that there is a relationship.
Mark

Mark,

I'm very sorry to say this, but I don't see any way that a 54/67 match could exist for people who are related within only 2 generations. Mutations simply don't happen that quickly. They aren't caused by alcoholism, illness or radiation -- these mutations are rare copying mistakes that occur at conception.

A single mutation between father and son is reasonable, and maybe two would even be reasonable. But mutations don't happen between every father/son pair, so they don't build up to the level you're looking at between close relatives. If mutations happened in every generation of a family, then Y-STR testing would be fairly useless for genealogy. The power of Y-STR testing is that we can test very distant cousins and, in most cases, expect them to match at 60/67 markers or better.

There is one particular mutation event that could cause seemingly higher mutations -- RecLOH. That's an abbreviation for "recombinational loss of heterozygosity". In layman's terms, it's a different kind of mutation that can cause a doubling of values at multi-copy markers such as DYS385, DYS459, DYS464, YCAII (any marker with a, b, c, d labels). Essentially, one copy of the marker gets overwritten with the value of another copy. So, for example, an 11-14 at DYS385 could become 11-11 or 14-14 in a single mutational event (rather than 3 separate 1-step events).

I don't think even a recLOH would cause the level of differences you are seeing, but let's check. Take a look at the two sets of results and see if there's any significant doubling at the multi-copy markers. If so, we can take a closer look. But if not, then I have to go back to my original statement that 54/67 isn't possible for two people who are supposed to be related within 2 generations.

You also asked how two unrelated people could randomly be seemingly close at 54/67. Well, if you are in haplogroup R1b, that is THE most common haplogroup in western Europe, and that level of mutations for unrelated (within genealogical timeframe) people is probably quite common.

If you want to share the haplotypes with us (best to upload to Ysearch and give us the IDs to look up), I'm sure someone here will be glad to give you additional analysis of each haplotype.

Elise

[capeflier]

Thanks

Thanks Elsie it and Jack- All but 3 of the markers that did not match on the 43 Marker test were tested at both FTDNA and the other lab. Again I am fairly comfortable that the results are correct but to be sure I will test my sons as well. Im not sure how the resemblance in the photos could come from a anything but Y side because except for the father the families would have never had a reason to be within 1000 miles of each other. There are 3 members of my family who look like three different members of the other family. It could just be cruel coincidence.

What I was looking to understand was that if you could have 2 mutations in the first 12 Markers or 2 in the second 12 and so on I didnt see why on a very rare situation it couldnt happen all at once and then there would be 11-13 mutations in 67 markers. As you describe it Elsie thats not the way it works, I dont have the ok to post the other parties results at the moment and my results alone probably wouldnt help much. We have a 2 marker mismatch on 464bd but otherwise no so I dont see how a RecLOH could have much of an effect on these results.


Thanks again for your input Id rather have your answer then no answer at all.
Mark