Different results between Ancestry and MyHeritage DNA tests

[Luke]

Apologies if this has been raised before ... I had a quick look through the topics but couldn't find a heading that matched.

I originally tested with Ancestry DNA, and since uploaded my results to MyHeritage (and GEDmatch).

Yesterday, I received an e-mail alert from MyHeritage, letting me know I had a match with someone from interstate. He was my sixth "best" match, with 47.6cM, giving an estimate of 3rd-5th cousin.

I messaged him to discuss the match and got a quick reply. We are still working on where we match though I have a good guess. I asked if he had an Ancestry DNA profile as well, which he did.

When I went to Ancestry DNA and found his profile, it determined that we only had 7cM of shared DNA, and he was a distant cousin. While MyHeritage brings up several shared DNA matches, a couple of whom have the surname where I believe we may be linked, Ancestry does not bring up any shared matches.

My question is, have any of you had a similar situation? Do Ancestry DNA and MyHeritage DNA test different DNA segments to have these significantly different results?

(PS: the MyHeritage DNA result's largest shared segment was 22.4cM)

 

[Pauline]

I raised a similar question back in 2017 which you can see here, and which may answer some of your questions.

Ancestry includes only 'close cousin' matches when we look at shared matches, and will not list those matches who are in our 'distant cousin list. While I can see that the list of shared matches could be impossibly long if all matches were included, it does mean potentially useful shared matches are omitted.

 

[Luke]

Many thanks Pauline!

 

[TerryM]

I think the big question in the reported results is the difference in relatedness in the two different sets of results.

Someone can correct me if I am wrong, but I believe that each company only analyses some of our DNA as the entire code would be too much for systems to handle. And we can conclude that each company has chosen different genes to analyse and hence the different results for the same two people's comparisons. This no doubt then would flow on to each company concluding different ethnicity as well.

And it means that if the number of base pairs analysed by these companies were to increase, our ability to find matches further back would be increased.

 

[Alexander Bisset]

The answer is down to Ancestry's Timber Algorithm. This algorithm looks at "common" DNA segments and discounts them as a match if they are too common ie: shared by hundreds of thousands of people. This results in Ancestry often having lower cM numbers than other sites. It can also lead to oddities where someone has a small break in there DNA they share with a parent and as a result they match with someone at a higher level than their parent as the algorithm knocks out the bigger number the parent has and leaves the parent on a smaller match number. Something that simply shouldn't be possible and wouldn't be without this Timber Algorithm.

Note Ancestry's argument is that it's pointless chasing down matches that hundreds of thousands of people have so by eliminating these common elements you can better find matches.

 

[Helen7]

It can also lead to oddities where someone has a small break in there DNA they share with a parent and as a result they match with someone at a higher level than their parent as the algorithm knocks out the bigger number the parent has and leaves the parent on a smaller match number.

Thanks, that explains why I have a match of 45 cM across 2 segments and my son has a match of 50 cM across 3 segments with the same person.

 

[Helen7]

This results in Ancestry often having lower cM numbers than other sites.

I find the cM are always significantly lower at Ancestry than other sites. I have a match of 94 cM at MyHeritage, whereas the same person is 66 cM at Ancestry.

 

[Bryman]

It can also lead to oddities where someone has a small break in there DNA they share with a parent and as a result they match with someone at a higher level than their parent as the algorithm knocks out the bigger number the parent has and leaves the parent on a smaller match number.

That is very useful information Alexander. I have matches with a mother and son combination where the son matches with me for 43 cM over 3 segs but the mother is only 35 cM over 2 segs. I had assumed that must mean that the father is also related to me but had not tested. I have had a lot of difficulty trying to find out how he is related and now realize that he may not be after all and that I could be wasting my time. Thank you.

 

[Alexander Bisset]

I have matches with a mother and son combination where the son matches with me for 43 cM over 3 segs but the mother is only 35 cM over 2 segs. I had assumed that must mean that the father is also related to me but had not tested. I have had a lot of difficulty trying to find out how he is related and now realize that he may not be after all and that I could be wasting my time. Thank you.

35 & 43cM are high enough that they will be related. As ever it's often difficult to work out how.

 

[peter]

You can read all about Ancestry's TIMBER algorithm here

Something that many people who test their DNA may not be aware of is that because we have two sets of autosomes and the tests generally can't differentiate between them, something that shows up as a matching segment could have been inherited partly from your father and partly from your mother (the same applies to the person you've been matched with). The phasing algorithm that Ancestry use is designed to compensate for this, but it is not and could never be perfect, because it depends on statistical inference.

(This is one reason why testing at least one parent is advisable, if you;re lucky enough to have a parent still alive.)

When the phasing algorithm gets it wrong it can result in a match with a child that doesn't show up as a match with their parent (assuming that parent has also tested). This can also happen as a result of an error in the test - low-cost consumer DNA tests aren't 100% accurate (though they're sufficiently accurate for our purposes).

You can read more about phasing on the ISOGG site.