Between Full and Half: Three-quarter Siblings


I’ve made an error in some of the analyses for this post.  I will update it as soon as I correct the errors.  In the meantime, I have withdrawn the post. 

Stay tuned!

15 thoughts on “Between Full and Half: Three-quarter Siblings”

  1. My grandma’s dad died (poisoned, according to my father!) about 1865. Her mother, after a while, married his nephew (almost as old as she was – and the family, according to my father, thought this was a good way to take care of her!), son of his brother. They had three kids. Years go by. A DNA match of mine wants to know who made his grandma pregnant . . . about the time period the kids of the kid would be impregnating people (maybe) and in the area (well, in an urban area nearby – in Raleigh, NC). So . . . turns out the three boys of the kid of the nephew were candidates . . . (one was too young). The two candidates have no surviving offspring we know of. We will never know which of the two . . . but it sure did skew the DNA! (We discuss going to the graveyard at midnight with a shovel, but haven’t gotten quite that motivated! )

    1. WOW! Fascinating story! There’s actually some interesting biological evidence that remarrying to an in-law is a good solution. The new spouse has an extra incentive to look after the children of the first marriage because they’re biologically related.

  2. Thank you so much for this Leah. So much useful information. I hope this will help others in their DNA journey

  3. > Half siblings are only related through one parent, so they shouldn’t have any FIR

    There’s no reason to expect that IBS won’t come into play, so you COULD have false-positive FIR.

    1. True! Endogamy can also contribute to spurious FIRs. Those FIRs will be small, so they should be easy to identify. I don’t think I mentioned this in the post, but three-quarter sibs should have at least 7 FIRs and generally more than 15.

    2. True! Endogamy can also contribute to spurious FIRs. Those FIRs will be small, so they should be easy to identify. I don’t think I mentioned this in the post, but three-quarter sibs should have at least 7 FIRs and generally more than 15.

  4. The descendants of 3/4 siblings are related as half cousins from the generation to the spouse who married two siblings and as full cousins one generation back to the parents of the siblings. I have a distant cousin whose relation is H4C and 5C. If you’ve also got endogamy, there are even more relationships.

    The correlation of relationship captures this and can also be used to estimate the expected cM. The problem is, there is empirically NOT a linear relationship between COR and shared cM (N=129 in my study with 0.0000015<COR<0.51). The reason for this is unclear. It could be sticky segments (perhaps topogolicaly associating domains=TADs), selection biases, segment imputation issues, or other reasons.

    Are your simulations based on segregation conforming to a normal distribution. Most of biology does not obey this assumption. You are also presenting a case study with an N=1. The discussion is interesting, but the conclusion need more supporting data and some consideration of the statistical assumptions. Given your careful work, I'd guess you're aware of this and working on more clarifying data! My purpose in commenting is to point out that we are in some gray areas where more knowledge is needed.

    1. Thank you for your comments. Can you explain what you mean by “segregation”? I know it as a phase of mitosis and meiosis, but you seem to be using the term differently.

      I will refer you to Caballero et al. (2019) for the assumptions underlying Ped-sim.

      What about the conclusion needs more data? In this case the answer is known; there is no doubt that the two women are three-quarter siblings.
      The goal of the post was to teach people about FIRs and to demonstrate how they can be used in a case where the total centimorgan amount is ambiguous as to relationship.

      1. I was referring to Mendel’s second law of inheritance: segregation. We now know crossovers are the mechanism by which genes segregate into two different daughter chromosomes. Mendel hypothesized that this was random. But we know now that this is not always the case. TADs are a notable example, as are the differences between females and males in the paper you cited. We know there are several other deviations from Mendel’s laws: anticipation, parental effects and maternal inheritance. The TADs are the most recent exciting exception and these may be quite helpful in genetic genealogy as “sticky segments” for probing more distant ancestors?

        You have actual FIR results on one match-pair. Yes, their relationship is known. But we do not know from actual testers whether the simulated distribution reflects what is actually happening in a large population. The test of this would be getting more ¾ sibs tested.

        The Caballero paper discusses the assumptions and does NOT use normal distribution, but rather a Gamma distribution. It’s implemented with Poisson methods. Thanks for pointing me to the methods paper!

        I appreciated the messages you communicated in your informative post. My point was that things MAY look different with a larger set of data on ¾ siblimgs. We need to keep open minds in gray areas were the data is slim.

        1. I think you are conflating two of Mendel’s laws. The law of segregation states that offspring get only one of the parent’s two alleles. That happens even in the absence of crossing over. The law of independent assortment holds that the inheritance of one trait doesn’t affect the inheritance of another. That effect is due, in part, to crossing over as well as the fact that there are multiple chromosomes. Independent assortment fails when two genes are positioned near one another on the same chromosome, something Mendel couldn’t explain but that is integral to 23andMe’s research model.

          I don’t think TADs are relevant here. They’re small—much smaller than the 7-cM threshold where we generally work—so even if they do have reduced crossover rates (Do they?), it wouldn’t show up in our matching. Even if TADs don’t cross over at all, they still segregate normally.

          Caballero et al. validated their approach using two different empirical datasets. Do you have reason to believe that DNA inheritance would be different in three-quarter siblings than in the relationships they used?

          That said, I made an error tallying segments from Ped-sim (My workflow is a bit clunky!), so I’ve retracted the post until I can figure out how to fix it. I may have to switch to ibdsim2.

  5. One way to deal with the potential endogamy and the way it raises HIR segment counts is to raise the threshold for a matching segment. You don’t say what the comparison at Gedmatch uses, but presumably the default of 7 cM. Raising that to 10 or 15 cM will give a more accurate count of segments shared due to recent common ancestry. The PedSim simulations would also need to be rerun with the same minimum segment size. There will be a trade-off between showing the correct relationship as a possibility and the fraction of the cM and segment ranges that overlap for the possible relationships.

    1. Thanks for your comment. It’s especially valuable because while tweaking the thresholds, I realized I did something wrong in tallying segments! I haven’t quite figured out how to fix it yet, so I’ve withdrawn the post for now.

  6. How might this show in children of 3/4 siblings? I have a “first cousin” with whom I share 498 cM on 21 segments. I also have a “second cousin” on the same side (my mom’s) with whom I share 336 cM on 11 segments. Since I’ve gotten the results I’ve wondered if my mother and her sister had fathers that were brothers. Everyone from my mothers generation has passed, so testing in my generation is the only option.

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.