Wednesday, April 30, 2014

Should I Test My Siblings Also?

A few days ago I got an email from a colleague in California who raised the following issue:

I suggested my brothers each take the AncestryDNA test while its on sale for $79.  One just wrote, "Sis- Didn't you already do this? If so, wouldn't yours and mine be the same?" 

Can any of you give me a quick answer off the top of your head to his question?   

I recall a lecture given by Katherine Borges last year. She showed us that her brother's results WERE different than hers, but I can't remember if it's something that an AncestryDNA autosomal test would catch.

"Celibate" vs "Promiscuous" DNA 

Some of your DNA is passed down to you from one of your parents without recombining with DNA from the other parent. I sometimes call this "celibate DNA." Since it is inherited intact, the information contained in it is identical or almost identical generation after generation. Unless there is a non-parental event (NPE), brothers will share essentially the same yDNA passed down from the father, and all siblings will share essentially the same mtDNA passed down from the mother. Therefore, unless parentage is in question, testing a sibling may be a bit of overkill. Currently in the US marketplace FTDNA is only major company testing celibate DNA for genealogists. 

Other parts of your DNA are recombined with each intergenerational transfer. By a way of contrast I call this "promiscuous DNA." As children inherit promiscuous DNA they get a different information package which is a mix of the contribution of each parent. Some of the information from each parent is lost in the process and some is retained. Into this category I place atDNA and xDNA. 

Autosomal DNA (atDNA)

Three players are active in testing atDNA in the US: 23andMe, AncestryDNA, and FTDNA. Any test you order from the first two will be atDNA tests as will the Family Finder test from FTDNA. Since atDNA is recombined as it is passed down to each child, siblings DO NOT inherit exactly or even nearly exactly the same instruction set unless they are identical twins.

On average the segments of atDNA inherited by two siblings will have a 50% overlap. The range of overlap can be as great as 63% or as low as 37%, but most will cluster near 50%. While this distribution accounts for differing phenotypes among siblings, such as hair color, etc., it also has implications for genetic genealogy.

Two (or more) siblings will not necessarily be shown as matches with the same individuals in a genetic genealogy database. Basically, as more siblings get atDNA results, more DNA matches will be discovered. The inheritance pattern of atDNA is random but there are statistical averages that guide our understanding. You can expect that you will match any close relatives who have also tested and are in the same database. As you look farther back for more distant relatives, the probabilities narrow significantly. You can expect to discover matches with:

  • almost all -- greater than 99% of biological 2nd cousins or those even closer related;
  • about 90% of those who are 3rd cousins;
  • about 50% of 4th cousins;
  • about 10% to 12% of 5th cousins; and
  • less than 2% of more distant relatives.
You will get at lot of very distant cousin matches because your distant ancestors have had so many generations to pass their DNA down to their present day descendants.

Testing Multiple Siblings

The same probabilities of matching cousins who are in a database will apply to each of your siblings who take an atDNA test. However, their probabilities operate independent of yours. For relatives who are 2nd cousins or closer, each sibling will have a high probability of having identical or near identical lists of matches. Each of you will have a greater than 99% probability of matching with any close relatives who are in the database. Testing additional siblings will not do much to enhance your number of matches for very close matches.

My background in math is somewhat in need of verification. As an undergraduate history major, I did take College Algebra. However that is as far as my math education went. I would welcome you readers to correct what I am about to set forth below. I have a feeling that I have the right idea but that I may be leaving out a critical step that will throw off the results.

As you start to look for more distant cousins, the advantages of testing siblings become more evident. At the 3rd cousin level each of you will match about 90% of the actual 3rd cousins in that database. Two siblings would both match about 81%. However, one or another of two siblings would match about 99% of the actual tested cousins in the database. Each would match about 90% of the 10% not matched by their sibling. This would equate to an additional 9% of the cousins in the database for a total of 99%.

At the 4th cousin level each sibling will match about 50% of the actual 4th cousins in the database. Two siblings would match a cumulative total of about 75% of the possible cousins (50% plus 50% of the remaining 50%). A third sibling would raise the cumulative total to about 87.5% (75% plus 50% of the remaining 25%) and a fourth sibling to about 93.75%.

At the 5th cousin level the testing of additional siblings also helps raise the match rate from about 10% for one individual tested to about 19% for two siblings tested (10% plus 10% of the remaining 90%), to about 27.1% for three siblings tested (19% plus 10% of the remaining 81%), to about 34.4% for four siblings tested (27.1% plus 10% of the remaining 72.9%), and so forth.

These numbers are at best approximations. Again, I would appreciate it if some of you could correct any computational errors contained above. However, the clear bottom line is that a much higher number of your actual 5th cousins in an atDNA database -- perhaps more than 3 times as many will be identified if you and three siblings test than would be found if you alone were tested. 

Other advantages of testing the atDNA of as many siblings as possible build on this kind of scale and allow an ambitious genetic genealogist to begin to trace which slice of atDNA came from which relative (chromosome mapping) and other advanced analysis techniques.

Blaine Bettinger points out that if you have been able to test both parents the value of testing all your siblings may be diminished. Especially when examining atDNA, testing the oldest generation possible is always the most desirable. There are fewer recombinations to dilute the information we can find.


  1. I agree with your conclusion that you should test as many siblings as you can, but I think there is a flaw in your math. Your multiplication of the percentages for each additional sibling would be correct if they were independent events. But they are not 'random' (say) fifth cousins of the mark, they are all children of the same fourth cousin once removed. That fourth cousin once removed has an approximate 25% chance of sharing DNA with the other person. The more children of that fourth cousin you test, the closer their combined chance of matching the person will be to 25%, but it can never exceed 25% because the DNA would have to have come through their parent.

  2. This blog is so nice to me. I will continue to come here again and again. Visit my link as well. Good luck