“Gordon” (not his real name) is an adoptee who recently contacted me to help decipher a new DNA match named Rafael. Another genealogist had previously identified Gordon’s birth mother. Based on her family tree, Rafael should have been a first cousin once removed (1C1R) through Gordon’s birth mother, but the amount of shared DNA was unusually high at 762 cM. Another maternal 1C1R, Beryl, shared 997 cM, which is beyond the known range for that relationship. Something unexpected was clearly afoot.

My first thought was that Gordon’s parents were related to one another. I ran his DNA results through the Are Your Parents Related (AYPR, for short) tool at GEDmatch, and indeed they were. Closely.


The ‘Are Your Parents Related’ Tool

AYPR looks for so-called runs of homozygosity (ROH), which is a geeky way of saying ‘sections of DNA that were inherited from both parents’. ROH are not uncommon in people from endogamous populations, where spouses are often distant cousins to one another. For example, my mother is Cajun, and this is a portion of her results:


She has a single ROH of 8.3 cM. It’s the solid yellow chunk on chromosome 2 in the image above.  Kitty Cooper shared a great rule of thumb in a recent blog post about AYPR: multiply the total amount of ROH by four to get an estimate of how closely related the parents are. In my mom’s case, 8.3 x 4 = 33.2. Her parents were, in fact, fourth cousins to one another, which fits the amount of ROH pretty well.

That’s not what was going on with Gordon, though. His AYPR results look like this:


All told, he had 20 ROH segments totaling 791.9 cM.  Using Kitty’s multiplier, 791.9 x 4 = 3167.6 cM, suggesting that his parents were full siblings to one another.  Or father–daughter.  It was a shock.

Gordon is a man with intrinsic grace, and he took the news well.  He feels blessed to belong to his adoptive family.  He is healthy.  And these results explained the message relayed by the adoption agency: His birth mother declined contact with Gordon because the experience was too painful for her.

After some time to process, Gordon decided that he wanted to know who his birth father was.  And since he did not want to intrude on his birth mother to ask, we took another approach.


Gordon’s Biological Family

Gordon’s biological mother was Helene Mills, born in 1942.  Helene’s parents were Oscar Mills and Florence Mattieson.  Florence died when Helene was only 7 years old.

Helene has two older brothers, Chris and Tony, and one younger brother, Michael.  Grandchildren of two of her siblings have tested; those are the matches that first tipped me off that Gordon’s parents might be related.  The numbers below Rafael’s and Beryl’s names indicate how much DNA they share with Gordon (total cM / number of segments), and the color indicates the testing company (green for AncestryDNA; red for FTDNA, with segments below 7 cM excluded).



From a purely genetic standpoint, there are four possibilities for Gordon’s biological father: Helene’s father Oscar or one of her brothers, Chris, Tony, or Michael.  We can eliminate two of those brothers as candidates based on other evidence, though.

Gordon was probably conceived in December 1957, when Helene was 15 years old and Michael was only 13.  Michael is unlikely to be Gordon’s father given his age at the time.

Rafael is Chris’ grandson, and he shares 762 cM with Gordon.  If Chris were Gordon’s father, Rafael would be Gordon’s half nephew through Chris (average, ≈850 cM) plus Gordon’s first cousin once remove through Helene (average, ≈425 cM).  If this were the case, Gordon and Rafael would share much more DNA than they do.  Thus, we can rule Chris out, too.

Only Oscar and Tony remain as possible candidates to be Gordon’s birth father.

I considered three factors in answering Gordon’s question:

  1. The number of ROH Gordon has.
  2. The total amounts of DNA Gordon shares with his DNA matches.
  3. Whether Gordon’s ROH segments are shared with paternal relatives, maternal ones, or both.

Number of ROH

Our DNA is packaged into units called chromosomes. We have 22 pairs of so-called autosomal chromosomes, plus one set of sex-determining chromosomes, the X and the Y. (Women have two copies of the X chromosome, while men have one X and one Y.)

A child inherits one copy of autosomal DNA from each parent, for a total of two copies of each one. Each copy in a pair contains the same sets of genes, but usually different versions of those genes.  That is, the copy of chromosome 3 from the mother is not genetically identical to the copy of chromosome 3 from the father.  (Think of gene versions like recipes:  your apple pie recipe might be different from mine, but they both make apple pie.)

When close relatives have a child together, some segments of DNA will be genetically identical on both copies. These are the ROH we saw in Gordon’s AYPR results above.  The total of ROH is a measure of how closely Gordon is related to himself through his two parents.  Unfortunately, the total cannot distinguish between a father–daughter scenario and a brother–sister one.

With every generational step, DNA segments along each chromosome get broken down in size in a process called “crossing over”. Because there are more generational steps in the brother–sister scenario than in the father–daughter one, we would expect the former to result in more ROH segments on average than the latter, regardless of the total amount of shared DNA.  If we know how many segments of DNA to expect in each case, we can compare Gordon’s actual number of ROH segments (20) to the expectations.

Dr. Andrew Millard is a professor at Durham University in the UK. He has developed a computer program that can simulate genealogical scenarios, and he generously agreed to do some simulations for Gordon’s situation. Simulations allow us to compare Gordon’s results to thousands of similar cases, even when we don’t have access to that many real-life data points.

First, consider the total amount of ROH and the number of ROH segments for the child of a father–daughter pairing versus a brother–sister one. The simulated data shows that the total amount of ROH in centimorgans is essentially indistinguishable in the two cases (left-hand image), but a father–daughter pairing will results in fewer ROH segments, on average (right-hand image).


Although there is a lot of overlap between the two scenarios on the right, when there are fewer ROH segments, the evidence weighs more heavily toward father–daughter, and when there are more ROH segments, it weighs more heavily toward brother–sister.

The blue arrow in the right-hand figure points to 20 ROH segments, the number in Gordon’s results. The figure suggests that it’s about 13 times more likely that Gordon’s parents were father–daughter than brother–sister.


DNA Shared with Other Matches

We can also consider how much DNA Gordon would be expected to share with his genetic cousins under each scenario. The figure below shows Gordon’s relationships to his closest known DNA matches.


Gordon’s direct line is in black bold text, and the people in red font appear twice in the tree. (Debbie is related to Gordon through both his Mills/Walton side and his Mattieson side).  As before, the shading indicates where the match tested.

If Tony is Gordon’s father, the matches on the Mattieson side are double cousins rather than single cousins, because Gordon would be the grandchild of Florence Mattieson twice. On the other hand, if Oscar is Gordon’s father, Gordon’s DNA should reflect the fact that Oscar is more closely related to the matches on the Walton/Mills side than Tony is.  The distinctions are subtle, however, which is why we must use simulations and evaluate all of Gordon’s matches as a whole.

For this analysis, Dr. Millard performed a bespoke simulation based on the relationships outlined above and then plotted the expected distributions of shared DNA amount for each of Gordon’s matches.  As before, we can compare the real numbers (shown with blue arrows in the figures below) to the expected distributions to get a sense of which scenario is more likely.

Two of Gordon’s matches, Beryl (997 cM) and Rafael (762 cM), are grandchildren of Helene’s siblings. If Tony is Gordon’s father, each of them is a double first cousin once removed, once through Helene (average, ≈425 cM) and once through Tony (average, ≈425 cM). If Oscar is Gordon’s father, each of them would be a first cousin once removed through Helene (average, ≈425 cM) and a half great nephew/niece through Oscar (average, ≈425 cM). In both scenarios, the average is expected to be roughly 850 cM (425 + 425) of shared DNA.


The simulations confirm that the shared DNA amounts are the same regardless of which scenario applies; neither Rafael nor Beryl helps to answer Gordon’s question.

For Gordon’s other matches, the amounts of shared DNA have different distributions under the two scenarios. Although the distributions overlap, in each case, there are centimorgan amounts for which the father–daughter scenario is more likely (where the solid green line is higher than the dashed red one) and others for which the brother–sister scenario is more likely (vice versa).


In Gordon’s case, every one of these cousin matches shares a DNA amount that favors the father–daughter possibility. Dr. Millard calculated that it’s 85 times more likely that Gordon’s parents were father–daughter than brother–sister based on these DNA matches.


Mapping of ROH Segments

Runs of homozygosity are segments of DNA that are inherited from both parents who, in turn, inherited the segments from their parents. If Tony is Gordon’s father, roughly half of Gordon’s ROH should be segments that both Helene and Tony inherited from their father Oscar and roughly half should be segments that they both inherited from their mother Florence. In that case, we should be able to find relatives on Florence’s side who match Gordon on those ROH segments. If Oscar is Gordon’s father, none of Gordon’s ROH should map to Mattieson cousins (although some non-ROH regions will map to those relatives).

Of Gordon’s 20 ROH, I was able to map five to known relatives. All of them are related to Gordon on the Mills side. The odds of this happening by chance are only 3.125%, indicating that it’s 32 times more likely that Oscar was Gordon’s father than Tony.



I analyzed Gordon’s DNA results three different ways, and all three ways supported the scenario that Oscar was Gordon’s father. The father–daughter pairing was 13 times more likely given the number of ROH segments, 85 times more likely considering how much DNA Gordon shares with his cousin matches, and 32 times more likely when ROH segments were mapped to the ancestors Gordon share with Gordon’s DNA matches. Overall, it’s about 35,000 times more likely that Oscar was Gordon’s father than Tony.


Implications for Physical and Mental Health

Discovering that your parents were closely related can be a shock, even though the child of such a pairing is never to blame. This brochure, written specifically for people who discover they have high levels of ROH, refers you to resources that may be helpful. https://www.watersheddna.com/highrohinfosheet

Runs of homozygosity can sometimes lead to medical conditions. Normally, we have two variant copies of each gene, so if one copy is defective or less functional than average, the other copy can usually compensate, and the individual is healthy. Within ROHs, both copies of each gene are identical, so if one copy is defective, so is the other. There is no “backup”. Of course, it is also possible that both copies will be fine and the person will have no adverse effects.

Having high levels of ROH is not a guarantee of medical problems; it just increases the chances. Your doctor will be able to advise you if you have concerns.

Because ROHs only occur when a child’s parents are related, medical conditions that arise are not likely to be passed on to the next generation. In other words, Gordon’s own children and grandchildren will not have an increased risk of genetic conditions as long as Gordon’s spouse is not related to him. While Gordon may have two identical versions of Gene A, Gordon’s children will have inherited one version of Gene A from him and one from his wife, so even if his copies are faulty, the one they inherited from their mother can compensate.

If you discover that you have high ROH levels and have additional questions, you may wish to consult a genetic counselor. This brochure can direct you to resources that will help.



Gordon’s story was shared with his permission. To protect the family’s privacy, all given names were changed to tropical cyclone names from the U.S. National Hurricane Center, and the surnames were made up.  Dr. Millard generously contributed his time and expertise to the case and pointed out that ROH could be mapped to cousins as evidence for who Gordon’s father was.

24 thoughts on “Gordon”

  1. Thank you so much.

    I have a very similar case (I’m working) but with much closer matches, including BM (deceased) full sister and MANY first cousins from both sides of the family.

    Is this process transferable to the case I’m working?

    1. If you’re trying to decide between father–daughter and brother–sister, you can count up the number of segments and compare to the simulated data from Dr. Millard. (You’ll have to count up the segments manually, because sometimes GEDmatch will artificially break up a single long segment into two or three smaller ones. Ignore anything less than 7 cM.) You can also use the strategy of mapping the ROH segments to cousins on the grandmother’s side. The other simulations were bespoke, meaning that they were custom done for Gordon’s family tree.

      1. Thank you!
        My husband (an engineer) is building a spreadsheet, using the adult child and first cousin from maternal side. I’m going to contact a first cousin from paternal side and see if he would be willing to update to GedMatch, IF you think that would be helpful.
        Thanks again.

        1. If you suspect that the parents were related, then yes, I recommend running the AYPR tool at GEDmatch.

        2. I ran the AYPR tool many months ago. It was confirmed already that it’s either her father or her brother.
          Once the results of a full sibling to the deceased BM came in, I’ve been trying to get help with this. I’ve not had the tools to resolve this mystery.

      2. The real complication with Gordon’s tree was Debbie who is related through both Oscar and Florence, but more remotely through Oscar. It sounds like the two sides in your case are separate and if you have lots of first cousin matches on both sides it should be relatively easy to map the ROH to one side or the other.

  2. Thank you Leah for publishing this. We covered this topic in the Advanced DNA class I just finished at GRIP. It’s terrific to see a real life example, and the steps you took to come to your conclusion.

  3. Wow! I have some surprises on my family tree, but nothing like that. As troubling as that is, I’m sure he has decided as we did that what is, is, and since you can’t change it, you might as well not worry about it.

    1. You’re absolutely right: Someone with high ROH bears no responsibility and shouldn’t feel guilty over their origins. Some people may need time and counseling to get to that acceptance, though. I would also encourage anyone with high levels of ROH to consult their doctor and possibly a genetic counselor, so potential problems can be addressed early.

      1. The surprises we got were not ROH relationships, but were definitely unexpected. In any case, it takes a while to come to terms with it, but probably longer in a case like this. I am not using my full name as we agreed not to tell generations younger than my children at this time, or anyone outside the immediate family.

  4. My husband has completed a spread sheet that compares the adoptee and a maternal first cousin.
    “I created a spreadsheet with the following information:
    – The adoptee’s ROH segments
    – The segments shared between the adoptee and a maternal first cousin

    I included only those chromosomes that contain both ROH and Shared segments.

    I then looked for matching segments between the ROH segments and the Shared Segments. For most chromosomes, there’s no matching segments. For some chromosomes, however, there are some segments that match. ”
    I’d love to be able to send you this (excel) spreadsheet (identities hidden) to review and tell us what your thoughts are. Is that possible?

    1. In Gordon’s case, I used Genome Mate Pro. I had Gordon’s matching segments from FTDNA and GEDmatch, and I tweaked the AYPR output so that I could import the ROH as if they were another match to him. That made it easy to find matching segments that overlapped the ROH.

  5. Yes, please. I need a second opinion and an understanding of what the threshold is for enough shared ROH segments to rule out BM’s father. The comparison is with a maternal 1C.

  6. Let me say this, to Lee it simple. There are ROH segments shared between the adoptee and his maternal 1C. Does this rule out the father? Thank you for your guidance.

    1. If the 1C matches the adoptee where the adoptee has ROH (and if those matching segments are sufficiently large, of course), then the adoptee’s maternal grandfather is not his/her father.

    2. We need to be very careful about terminology here. Leah and I had some issues of interpretation when we were discussing Gordon’s case. To whom is the ‘maternal 1C’ a 1C? What you need to compare are relatives of the mother’s father and relatives of the mother’s mother. If the mother’s father is the child’s father, then the ROH will only map to his relatives. If the mother’s brother is the child’s father, then the ROH will map to relatives of both the mother’s father and the mother’s mother. One ROH mapped to a mother’s mother’s relative indicates the brother, but you need multiple mappings to the mother’s father’s side to be sure that the mother’s father is the child’s father.

      1. Fair enough, Andrew. 🙂
        I’m reluctant to post much more in a public forum. Here goes:

        AYPR results were confirmed that adoptee’s (CM not cM) birth parents were first degree relatives in March 2018. A full sister to suspected BM was tested and shared 2,557 cM with CM. Two weeks ago CM received his original birth. cert. confirming the BM name. BF was listed as “unknown”.

        Shared ROH segments were from BM maternal line. I’ve requested a Ancestry tester from BM paternal 1C line to upload to GedMatch.

        There are many first cousin matches to CM on BM maternal line.
        BM maternal 1C DNA matches to adoptee are as follows (cM):
        1. 862
        2. 780
        3. 707
        4. 702
        5. 681
        6. 598
        These 1C matches came from every BM maternal branch siblings children.
        The BM paternal 1C match with CM at 652 cM.
        All family trees are well documented. Also, these cousins know each other and have reunions on a regular basis.

        Only surviving male siblings to BM are both far too young to be possible candidates as BF. The only older brother was 18 years old to BM 15 years old at time of adoptees conception.

        BM father was a fraternal twin. Adoptee was born a twin. His twin died at 2 1/2 months old in a foster home. Twin’s death cert. lists birth date is identical to CM. Twin has same surname as BM. Twin’s death cert. lists mother “unknown”. Informant for twin’s death cert. was listed as foster parent. Adoptee, CM was also in a foster home before being adopted. It’s unknown if adoptee was in same foster home.

        Additionally there is another adoptee (RL) who’s the child of one of the surviving male siblings to CM’s BM. RL is a female that I’ve been able to help her find and reunite with her BM. This adoptee shares 1,300 cM with CM.

        I hope that is enough information.
        Thank you for any assistance on this very sensitive subject.

        1. Julie, I suggest you email me and we take this discussion out of the public eye. My gmail.com address is millardnz.

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.