Revisiting the "Are Your Parents Related Tool" at GEDmatch

Most cultures have taboos against close relatives marrying (or mating with) one another. While some royal lineages historically used intra-family marriages to consolidate power, there are generally consequences. The practice is known to increase the risk of genetic problems in offspring.

That’s because most genetic disorders only manifest when the individual has two copies of a “faulty” gene, one inherited from each parent. The more closely related the parents, the more likely that is to happen. Genetic problems aren’t guaranteed, but anyone who is the product of a close-family mating should know their background so they can discuss it with their doctor.

Runs of Homozygosity

We inherit two copies of each autosomal chromosome, one from each parent. When two parents come from different backgrounds, the chance that their child will inherit two copies of the same DNA segment on any given chromosome is extremely small. However, if the parents were related to one another, the child may inherit segments that are identical on the maternal and paternal copy of the same chromosome. These segments are called “runs of homozygosity” (abbreviated ROH). The more closely related the parents, the more ROH the child will have, and the longer they will be.

ROHs occur in a range of situations. We see them in endogamous populations, where everyone is distantly related; when the parents were cousins of varying degrees; and in cases of incest, where the parents were closely related.

The “Are Your Parents Related” Tool

The easiest way to check for ROHs is with the “Are Your Parents Related” tool (AYPR) at GEDmatch. This tool compares each chromosome pair (the “mom” copy to the “dad” copy) and graphically displays the results as a horizontal bar with two tiers. The upper tier depicts the actual DNA comparison while the lower tier shows whether GEDmatch’s algorithm counts each region as an ROH.

When the parents are not related, the upper tier will be mostly red with occasional green “flecks” indicating tiny DNA that matches through random chance. These little flecks are not indicative of parental relationship; we all have them. The bottom tier in such a case will be solid grey, meaning GEDmatch’s algorithm doesn’t consider the chromosome to have any ROHs.

Occasionally, the green in the top tier of the graphic will be more than just a fleck, indicating the presence of an ROH. If the bottom bar remains grey, the ROH is too small to meet GEDmatch’s threshold and is not counted. These are usually a sign that the parents were distantly related or perhaps from the same ethnic background.

Finally, there are ROH segments that are 7 cM or more, meeting GEDmatch’s threshold for inclusion. These have green in the top tier and blue in the bottom tier. Depending on their size and number, these ROH could suggest anything from endogamy to a cousin marriage to incest.

Dealing with Difficult DNA

In her “Dealing with Difficult DNA” session at the Ohio Genealogical Society conference in May, 2025, Kate Penney Howard shared a side-by-side comparison of AYPR results for five different people.

Person 1 has neither endogamy nor recent pedigree collapse.
Person 2 has some endogamy.
Person 3’s parents are 4th cousins to one another.
Person 4’s parents are 1st cousins to one another.
Person 5 is the product of parent–child incest.

Do you notice something peculiar in these graphics? I did. Person 5 has several large ROH that are not being counted by the algorithm. That is, those segments are green on top but grey on the bottom.

Something was wrong with the algorithm.

The Glitch in the Machine

Kate and I both immediately realized the implications: DNA testers in highly sensitive situations could be getting incorrect information about their family backgrounds when the truth is essential for both emotional and medical reasons.

I’d already seen the problem first-hand. Several years ago, I helped “Gordon” with his DNA results. A previous genealogist had told him that his parents were second cousins, but I quickly realized that they were much more closely related. (In fact, they were father and daughter.) In Gordon’s case, his original DNA results were poor quality, so AYPR couldn’t analyze them properly. Garbage in, garbage out, as they say. When he tested with a different DNA company, AYPR gave clear results. Fortunately, Gordon is a man of immense grace, and he took the news well.

That’s not what was going on in Kate’s graphics, though; we’d seen the same thing with multiple people. This was clearly a problem with the GEDmatch algorithm.

Kate has provided feedback to GEDmatch about the AYPR tool in the past, so she volunteered to raise this newer issue with their tech team. They quickly identified the problem—an issue with how the algorithm interprets ROH at the ends of chromosomes—and promised a fix as soon as possible. That fix was implemented today.

Judy

Consider this (fabricated) example of an adoptee, Judy. Her original birth certificate lists her biological mother (BM) but no father. At the time of Judy’s conception, BM lived with her parents, brother, and paternal grandfather, while an uncle and male first cousin lived nearby. BM had no other living male relatives.

For Judy’s case, the AYPR tool (before the bug fix) reports an ROH total of 300 cM.

Kitty Cooper, a founder of genetic genealogy, recommends a straightforward approach to estimating how much DNA the parents shared: multiply the ROH total by 4. We can then use that product in the Shared cM Tool (SCT) at DNA Painter. This approach suggests that Judy’s parents shared roughly (300 x 4 =) 1,200 cM with one another, giving an 84% chance that Judy’s father was BM’s cousin, a 16% chance that it was either BM’s uncle or grandfather (≈8% each), and ruling out BM’s father and brother.

There’s another tool that can be used to interpret AYPR results, the “AYPR ciM” tool, which was recently integrated into the AYPR tool at GEDmatch. It gives very different probabilities but, like the Cooper + SCT approach, also definitively rules out both BM’s father and brother as Judy’s father. (This approach has not been independently validated with empirical case data.)

But what if the AYPR bug is underestimating the amount of ROH by, say, 20%. The actual ROH would be 375 cM. The Cooper–SCT method (using 375 x 4 = 1,500 cM) would predict a 99% chance that Judy’s father was either BM’s grandfather or uncle (≈49% each) and a 1% chance that he was BM’s cousin. BM’s father and brother are still ruled out. The AYPR ciM tool predicts 46% for BM’s grandfather, 25% for her uncle, 13% for her cousin, 0.2% for her father, and no chance that it was her brother. (Most numbers have been rounded to the nearest whole value.)

If the problem is even worse, and the AYPR tool is failing to tally, say 50% of the ROH, then the actual total would be 600 cM. The Cooper–SCT method gives a 99% chance that Judy’s father was BM’s brother, 1% that he was either BM’s uncle or grandfather (0.5% each), no chance that he was BM’s cousin. (A quirk of this approach is that it incorrectly rules out BM’s father, who is roughly as likely as her brother. This can mislead novices, but those experienced with ROH will recognize that BM’s father is about as likely as her brother.) AYPR ciM gives odds of 48% for BM’s grandfather, 23% for her uncle, 17% for her father, and 12% for her brother.

Taken together, these three scenarios highlight how drastically the bug in the AYPR tool might be affecting some of the most emotionally challenging cases we work as genetic genealogists. If people have used the AYPR total to determine the relationship between parents without taking additional safeguards—such as incorporating other matches using BanyanDNA or weighing information from those who may know of abuse within the family—they may have excluded the actual parent and incorrectly implicated a more distant relative.

The effect of the bug is likely to be trivial for more distantly related parents. For example, if the current version of AYPR reports 20 cM ROH and is off by 50%, the predictions are not much affected. However if, as in an example like Judy, the amount of ROH is high, any errors in estimation could have substantial effects.

Kate and I are both concerned that previously “solved” cases will require re-analysis and may have identified the wrong biological parent(s), especially if the conclusions relied too heavily on the AYPR output or the probabilities presented in the tool. If you or someone you know may be in this situation, please reach out privately and Kate or I can help you.