*SUCCESS* Autosomal DNA Test - my experience

[AngelaR]

That is most helpful, DevonCruwys, thank you 

I shall stop hogging Dudley's thread and go and explore.....

[DudleyWinchurch]

A quick update on the practicalities of the test. 

An email arrived overnight to let me know that the results were now available.

Having now reviewed my results several times, I am becoming concerned that I have already about six potential third cousins for whom I can find no evidence whatever for any possible relationship.  Not only no apparent match in names, but also families that appear to have been continents apart at the times when they would need to have overlapped! 

I understand that these may be matches due to anomalies in my chromosomes but, if there are so many of them, how does that affect the level of confidence for the test that I am hoping to make?

I was not expecting so many matches, but am now confused by finding a number of people flagged as relatively close kin that does not appear to be a practical possibility, whereas one or two of the more distant possible relationships do at least share a geographical location at an appropriate time in my family history.  If none of the supposed third cousins are likely, is it worth checking through the (hundreds of) others for a possible match?

[DevonCruwys]

The relationships are predictions and there will inevitably be a range because of the random way that DNA is inherited. We might receive 21% of our autosomal DNA from our paternal grandfather but 27% from our maternal grandmother. It's much more difficult to give a precise prediction beyond the second cousin level.

The other problem is that some of your matches might come from endogamous populations (where people married within their own community). As a result, relationships will appear closer than they actually because these people will have got a double dose of DNA from the same ancestral couple. This is a particular problem for members of the Ashkenazi Jewish community. It also affects Americans who have Colonial American heritage and you'll find that there are rather a lot of these in the database.

I find it best to ignore the American matches unless they share a rare surname and focus instead on the people who have ancestry from the UK. If you download the spreadsheet with your matches you can often identify the people living in the UK from their e-mail addresses (Virgin, BTinternet, addresses ending in .co.uk, etc). It's also worth following up with all the Australians, New Zealanders and Canadians as they are much more likely to have more recent documented links with the British Isles where you stand some chance of finding the connection.

[DudleyWinchurch]

Houston, we have a problem!  I seem to be plunged into technical depths trying to make sense of the findings.

Thanks Devon, for you explanations, but something further seem to have gone wrong here!

I think I have a problem with the difference between Identical by Descent (IBD) and Identical by State (IBS).  The suggestion seems to be that they somehow already filter for this (by not including strings that don't usually get mixed during the (mitosis/meiosis?? (long time since I studied DNA) process).  However, I think that maybe I have a number of unusual preserved blocks.

Of my 7 suggested third cousins, and one suggested 4th cousin, I have six where most of (up to 2/3 of) the match comes from a single very extended block.  Is that at all normal??

2 matches who seem to be related, share these very extended blocks in the same place on the same chromosome, the very extended blocks in common with another 4 are each on different chromosomes.

When I look at the comparison with my second cousin, we have several extended blocks in common, but none as long as the ones with these "matches", although in two cases (that are related to each other), I do share about half of the very extended block with my cousin.  I also have many other shared areas (looked at on the 1cM view of the chromosomes).

If these very extended shared blocks are not normal, is there some statistical method I could use to weed out false positives? - e.g. all those where more than half the shared data comes from one block?

Sorry, this has got so technical but I really don't know whether it is better to ignore pretty much all these potential matches or to follow up some of the ones that at least seem to be in the right place at the right time, even though not very close connections.

[DevonCruwys]

Dudley

IBD are true matches where there is a genealogical common ancestor. IBS are false positive or coincidental matches where by chance the same segment is shared at the population level. Some older segments get preserved purely by chance.

The segments that you are seeing in the chromosome browser are half-identical segments. The stop and start points that FTDNA give are not exact. You could be matching different people on different strands of the DNA. See:

http://www.familytreedna.com/faq/answers.aspx?id=17#844

These could also be compound segments:

http://www.familytreedna.com/faq/answers.aspx?id=17#612

I've tested myself and both of my parents. 18% of my matches do not appear in my parents' match lists and are therefore false positives, but these are all fifth to distant cousins.

The companies are not yet that good at filtering out IBS segments. It is estimated that 99% of segments that are over 10 centiMorgans in size are IBD, but when you get down to 7 cMs only 30% will be IBD:

http://www.isogg.org/wiki/Autosomal_DNA_statistics#Identical_By_Descent_segments

The techniques to establish whether segments are IBD or IBS involve chromosome mapping and phasing. Ideally you would need to test other family members. See these two ISOGG Wiki articles for more information:

http://www.isogg.org/wiki/Chromosome_mapping

http://www.isogg.org/wiki/Phasing

I would suggest your questions are best asked on the ISOGG DNA Newbie where there are lots of people who are much more experienced at interpreting autosomal DNA test results than me:

http://groups.yahoo.com/neo/groups/DNA-NEWBIE/info

[DudleyWinchurch]

Thanks Devon,

I have been trying to read up and understand all these links but it's very confusing and maybe I didn't make myself clear in my last message.

Most of these matches depend on a single longest block and very few other coinciding strings.

We're definitely talking over 15 centiMorgans and in some cases, over 30.  I cannot see how anyone could share 35 centiMorgans in one block, yet less than 15 in total elsewhere, unless due to some bizarre co-incidence.

My second cousin shares several extended blocks, the longest of which is just under 35 centiMorgans, and quite a few smaller ones, but this is from a total of 194 centiMorgans shared.  As I know this to be a genuine relationship, I assume that this is a fairly normal distribution pattern.

Yet there are two other matches, related to each other, which apparently share slightly more than 35 centiMorgans of that same part of the chromosome with me and yet share much less other matching DNA: in one case considerably less than half as many centiMorgans elsewhere.  I have been in touch with one of these and he can confirm no match with any of the names in my previous four generations of ancestors. 

A similar pattern, affecting different chromosomes can be seen with at least 4 other of these 3rd cousin matches, and none of them have any name or location information that suggests that we are related in any way.  Any information given would strongly suggest that we are not.

It looks to me that the only possible evidence for a DNA match here is a single (very) extended block on one chromosome and very little other evidence.

The fact that so many of these matches seem to follow the same pattern makes me wonder if there is something odd going on.  Do I have an exceptional number of archaic preserved DNA strings or something similar?

Whatever the reason, I don't think that they are true matches and was hoping someone might have a suggestion as to how to weed them out, in order to see if there are any true matches amongst this bizarrely long list.

[DevonCruwys]

If the matches have long blocks of over 15 cMs then they are probably all true matches. The size of the segments varies because the inheritance of autosomal DNA is a random process. Have a look at Tim Janzen's spreadsheets where he's tested lots of known relations to get an idea of the variations that are found:

http://mennodna.remotewebaccess.com/23andme/ref.html

If you can't find the documented relationship I wonder if your matches are perhaps Colonial Americans who have some cousins intermarrying in their family trees so that the relationships look closer than they actually are? Do they all have some geography in common?

The people on the DNA Newbie list will be able to give you better answers than me.

[DudleyWinchurch]

The problem is that not one of these matches is in long blocks.  Each has just ONE very long block and very few other bits and pieces (not one likely to be anyway near 10 centiMorgans) and, in more than one case, the long block is almost double the length of the total of the rest of the matches.

I can't see a scenario when nearly all of the shared DNA between relatively close relations would be packed into just one block.

The one that contacted me may well be a colonial American, pleased to have found a third cousin, but suggested that maybe we were related sometime around the Norman Conquest?

I have a lot of distant matches coming up with German or Scandinavian origins - none of which is likely within any of the documented family history that I have.  I also have one or two very distant ones that show possible family names or geographical locations that might make sense, but they are so distant compared with all these relatively close matches that don't make any sense at all that I am reluctant to bother following them up until I can work out what is happening with all these potential third and fourth cousins.

My brother is going to take the test as well so I can compare our mutual patterns.

[DevonCruwys]

If it's any help you might like to have a look at these Chromosome Browser example for known relationships up to the third cousin level to see if those patterns match yours:

http://www.isogg.org/wiki/Chromosome_Browser_Examples

Are your third cousin matches actual predicted third cousins? Mine at this level all have a range of anywhere between third and fifth cousins.

There does seem to be a phenomenon known as "sticky segments". Type that into the search engine of the Genealogy DNA list and read the related messages:

http://archiver.rootsweb.ancestry.com/cgi-bin/search?path=GENEALOGY-DNA