AncestryDNA Is Using FIRs to Distinguish Full and Half Siblings

Did you know that AncestryDNA is using fully-identical regions to distinguish full siblings from half siblings?  And apparently has been doing so since 2016?  I didn’t.

If you’re asking yourself, ‘How did we miss that?‘, I am to!  And I don’t have a good answer.  I didn’t read their scientific White Paper closely enough.  Mea culpa.

Then again, if you’re asking ‘Isn’t a FIR a tree?  The other kind of tree.  What does it have to do with DNA?‘, read on.

A fir, not to be confused with an FIR (any excuse for a good botany pic!):

Autosomal DNA and Relationships

Autosomal DNA, the kind most commonly used in genealogy, indicates how closely two people are related.  The more DNA they share, the closer the connection is likely to be.

It’s not so simple, though, as you can see from the graph below (modified from Figure 5.2 in the AncestryDNA Matching White Paper).  Each relationship has a range of possible shared DNA—some 2nd cousins share more than average and some share less—and the ranges of different relationships overlap.

The lower the amount of shared DNA, the more overlap there is among relationships.  Barring endogamy or other factors, a match of 500 cM could only fit one of three relationship categories (the groups that include 1C, 1C1R, or 2C, represented by the red, navy, and yellow lines, respectively), whereas at 200 cM, there are at least four possible categories:  1C1R/navy, 2C/yellow, 2C1R/light green, 3C/black, and possibly even 3C1R/light blue.

Even the closest relationships can fall into overlapping zones.  A half sibling (dark green line) who shares an above-average amount could match on more DNA than a full sibling (medium blue) who shares less than average.  There is a way to tell full and half siblings apart, though.

 

Fully Identical Regions

We have two copies of each autosomal chromosome (the ones that don’t determine sex).  For each chromosome, half siblings can only match one another on one of their two copies, the one they inherited from their shared parent.  They won’t match on the other copy, which they each inherited from a different parent.

Full siblings share two parents, though, so they can match on both copies of each chromosome.  They will have some sections of DNA where they inherited the same segment from their mother and, in the same spot, where they inherited the same segment from their father.  This is called a “fully-identical region” (FIR), and half siblings won’t have them.  Full siblings will also have regions where they only match through one parent, and some where they don’t match at all.

Genetic genealogists have long known that we can use FIRs to distinguish full versus half siblings, but we are only able to visualize these regions at 23andMe and GEDmatch.  Here’s a visual comparison of three chromosome in my two kids at GEDmatch:

The solid green regions are where they inherited matching DNA from both me and their dad, the mostly yellow regions are where they match through just one of us, and the mostly red regions are where they inherited different DNA from each parent.

Half siblings will have red and yellow sections when viewed at GEDmatch but no solid green (FIRs), like this:

 

FIRs at AncestryDNA

At AncestryDNA, matches are categorized into groups based on how much DNA is shared.  For example, a match who shares between 650 and 1300 cM will be in the “First Cousin” category (although other relationships are possible), a match between 200 and 650 cM is in the “Second Cousin” category, and so on.

Full siblings normally fall into the “Immediate Family” group and half siblings into the “Close Family” one.  If those categories were delineated solely by a centimorgan threshold, full siblings who share less than 2200 cM would be classified as “Close Family” and half siblings who share more than that would fall into the “Immediate Family” category.

In the past, that’s exactly what happened.  But it seems that AncestryDNA updated their algorithm in 2016 to consider FIRs.  The weird thing is that no one noticed!  Or if they did, they didn’t make a fuss over it.  I certainly missed it.

This figure, also from the AncestryDNA Matching White Paper, shows how the consideration of FIRs (labeled IBD2 on the horizontal x axis) helps to distinguish full (orange) and half (red) siblings, and even to tell identical twins (purple) from a parent or child (blue).

 

Considering only total shared centimorgans (the vertical axis), a match of 2200 cM could be either a full sibling or a half sibling.  Both red and orange points fall there.  However, when you plot FIRs (IBD2) on the x axis, the two relationships are distinct.

Parent–child and identical twin relationships can now be similarly differentiated.  Of course, most people with those matches never doubted the relationship once age was considered.

This change is a big improvement for close relationships.  That it flew under the radar for so long is pretty amazing!

Take home lesson:  AncestryDNA is doing a better job of telling full sibs from half sibs than they once did.

 

25 thoughts on “AncestryDNA Is Using FIRs to Distinguish Full and Half Siblings”

  1. Thangs so much for the blog. What about in the case of a man fathering children with women who are sisters of each other (not related to the man!), or a mother having children with men who are brothers (not related to the woman)?. If the sibling parents were very similar genetically, it seems this algorithm for distinguishing full vs half sibling might not work.

    1. The cases you are describing are sometimes called “three-quarter siblings”, who will definitely share FIRs, but not as many as full siblings. I haven’t seen a plot for 3/4 siblings, but I suspect Ancestry would categorize them as “Immediate Family” along with full siblings. It would be great to be able to tell them apart.

    2. I used Visual Phasing to help a woman determine if she and her 3 siblings were full or 3/4 siblings. FIR alone couldn’t tell, but comparing them in pairs in the Visual Phasing spreadsheet I could tell there were two closely related fathers.

  2. Huh. When I discovered in 2016 (at age 64) that I was adopted, the adoptee search community (and everyone else!) kept hammering the point that the only way to validate my matches was with a chromosome browser. Fine, being completely new to DNA for genealogy I took their advice. Having tested with Ancestry I also took their advice to transfer my results to GEDMatch, and to “fish in all of the ponds” so I also tested at FTDNA and 23andMe. I got nowhere fast because my closest matches were all on Ancestry and once you get a reply in Ancestry’s horrible messaging system (if you ever do), adoptees (aka total strangers) generally don’t have a lot of success with getting a match to transfer to GEDMatch.

    So, I just rolled with the flow and gave up on chromosome browsing. The clues I had already within my Ancestry matches seemed to validate themselves with the estimated relationships (in cM’s and degrees) and by validating their trees using Ancestry’s excellent record databases. I created spreadsheet matrices of my Shared Matches because I could clearly see at least two distinct Shared Match groups that indeed turned out to be my paternal and maternal lines. The matrices also showed other groupings that further split the two lines, and I was able to target specific branches by using those matches with public tree starting points for my record research.

    I was amazed at how well Ancestry’s much maligned Timber and Underdog (mostly Underdog) algorithms appeared to work for me and I shared my experience in multiple groups & forums. Not very many people listened.

    But two years later I knew who both of my birth parents are (Dad is deceased) and am enjoying conversations with my Mom and meeting my maternal and paternal half siblings, nieces & nephews, great nieces & nephews.

    Whatever it is that Ancestry does (now I know the final puzzle piece, so thank you!) has had me as a minority but solid believer since 2016. It is also why I’m skeptical that the “new” methods of match visualizations (colored Shared Match matrices) will result in similar insights with test results from companies other than Ancestry.

    I’d love to see you do an in depth article on the matching methods that My Heritage does since it’s the only other company I’m aware of that claims to do phasing.

    1. Congratulations on finding your birth parents, and especially on the positive reunion with your mother’s side. I agree with you about chromosome browsers: I rarely use them myself (only for really specialized scenarios), and I solve most of my unknown parentage cases with just Ancestry.

  3. Have recently been contacted by 1/2 a twin (male & female) born illegitimately like myself to whom we’ve been able to determine father’s identity via 1st cousin links. This party’s mother also confirmed father’s name when confronted. I am a bit confused as to how we link since we have different mother, albeit same father. Apparently I was contacted as 1st cousin but when comparing connection we determined we’re half siblings. Would the twins both carry similar genetic material to me (I’m the younger of them) or what?? I filled them in on what research I had done on my father’s line until my nephew and cousin plus 3rd cousins decided to hide the family tree from me. I wasn’t wanting connection but did want to know my biological father’s line for my tree. Was told same thing happened with them.

      1. I believe he said he was over the 1,000cM range. With me being female I am 868.8cM to which I was told of 1st cousin and gr. nephew that enabled me in determining my biological father. His sister is a bit lower than I and that was all the info he gave me. When he wrote he said I was cited 1st cousin to him but we were able to conclude the biological parent being our father.

        1. One thousand centimorgans is too low for a half sibling. Exactly how much DNA do you share with the male twin? How much do you share with the female twin?

  4. He didn’t volunteer much other than to say that I, and some parties mentioned in my match also matched he. (This included 3rd – 5th cousins but not a 2ndC1R as this party is on my mum’s side). He did say his cM count was higher and his sister’s lower to mine – I’m female and mine is 868.8 to my 1st cousin and gr. nephew. Within this match and 3rd – 5th cousin matches I was able to determine my biological father

  5. As I haven’t heard from party, as yet, I reread his email sent and it appears I wrote incorrectly. His count of over 1,000cM isn’t to me but the same party I have for 1st cousin and gr. nephew and likewise to my 3rd – 5th cousins; however, not all of the latter. My name was listed in amongst his ‘match’ listings on the second page within ‘high’ connections. His sister, other half, has a near 700cM and, I’m assuming, this means we are closer linked than I to her brother. Within his first set of matches/links, as said earlier, we have same majority of named parties (none of respective mothers although looking at his research there seems to be a link between us at gr. gr. uncle lineage – have to check more into this as may be coincidence of same names and spouses with circa years entered) all of whom are on my biological father’s lines. I was adopted whereas they were raised by their mother and step-father – I did reunite with my biological mother and family and was told name of father which appears confirmed by my DNA match of 1st cousin and great nephew – top of the listings.

    1. Sounds like you’re on the right path. If you’re interested in hiring me for a consultation to guide your next steps, please email me at theDNAgeek (a) gmail.com.

      1. Hate to ask this but for how much?? I am really trying to get a bead on my ancestry/heritage lines more for health reasons as it appears both sides of my lines don’t have long lives and am curious to know why. I’ve found cousins and aunts dying in their 50′, 60’s with only a few making it into 80 years. Earlier, like 1700 and early 1800 had better lifespans – I know many believe evolution brought many changes but feel there is more to it than that. New family I’ve reached out too aren’t interested in the past (citing they have enough to worry about with today/tomorrow) and/or would prefer no communication so am busy trying to determine my father’s side, in full, and acquire death certificates. I am in my 60’s and resources are limited, I’m afraid.

        1. I have a brother and sister who have tested. They share 2342 cM per Ancestry and are reported as full siblings. I suspect that they are 3/4 siblings sharing a mother with the fathers being brothers. So both uploaded to GEDmatch and the total cM of FIR’s is 408, which is much lower than the appx 850 that I was suspecting. Would this indicate 3/4 siblings? Thank you for your thoughts.

        2. It’s difficult to tell full from three-quarter siblings. Unfortunately, I do not have a set of distributions that could tell you for sure one way or the other.

  6. Interesting information. I have a question. Will an Autosomal DNA test confirm that me and my half sister are indeed sisters when we share the same Father? He acknowledges her and my half brother (her full brother) and our older half sister. But refuses to acknowledge me. Years ago my half brother was willing to take a DNA test but back then they were $400. Now he isn’t even speaking to his sister or mom do I doubt I could get him to take the test(or if it would even tell us anything). My sister just submitted her test but I am now worried it will not show any relation and that it will be a waste of $69.

    1. If she’s your half sister, you and she should match in the “Close Family” category at AncestryDNA and should share about 1800 centimorgans (could be more, could be less, but it’ll be in that range). One additional way to confirm that you share a father would be to both transfer your raw data files to GEDmatch and do a “One-to-One X-DNA Comparison”. Sisters who share a father should match across the entire X chromosome. (A little bit missing at the end is nothing to worry about … that’s just an artifact.)

  7. my sibling brother and I shared 2135 cm on family tree DNA and DNA Painter showed it to be about an 81.42% likelihood of 1/2 siblings – only 18.58% full sibling.
    I recently took an Ancestry test and this one says 2175 over 69 segments and says full sibling. I had a little reason to think we might be something like 3/4 because I had suspected we might have different fathers. But the fathers would have been related..as their grandfathers would have been brothers. In other words my brother “could” be about a 3rd cousin on the male side…any info or input would be greatly appreciated. We are older and parents are deceased.

    1. At Ancestry, click on the little “i” in a grey circle next to the amount of shared DNA. What does it say there about the percentage chance that you are full siblings or half siblings?

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.