No announcement yet.

comparing Family Finder with 23andMe

  • Filter
  • Time
  • Show
Clear All
new posts

  • comparing Family Finder with 23andMe

    Does anyone have any information about the accuracy of 23andMe versus Family Finder?

    I have a Family Finder match whose longest shared chromosome segment is on chromosome #15, 10.57 cM ranging from 21,487,876 to 25,774,171. The same person matches me at 23andMe with a long segment also on chromosome #15 but 32.98 cM extending from 1 to 29,327,352--a significant difference.

  • #2
    It may be due to either a difference in the algorithms used to interpret the chip results, or what the chip is programmed to recognize, or a combination of both. You can see some such differences shown in the Autosomal DNA testing comparison chart between the two companies.

    I tried to find some reference online that covered this situation, but the closest I thing I came across was from about four years ago, when CeCe Moore wrote about "Discrepancies with Amount of Shared DNA for Close Family Matches at MyHeritage," in which she gave several examples of discrepancies between total and largest matching segments reported at MyHeritage, when compared to results for the same people at GEDmatch or Ancestry. In those cases, it was an issue with how MyHeritage was calculating cM segments; the MyHeritage Chief Science Officer even made a comment to the post about how they were working to improve it. I know in your case a close relative is not involved, but perhaps we can see how one or another company's interpretation might be considered faulty, whether due to their algorithm or chip. I haven't seen anything discussing problems with either 23andMe or FTDNA specifically regarding accuracy, though.


    • #3
      Your match might be a moot point.

      . Chr 15 pileup.PNG

      You are in TWO pile-ups. So might not be a very solid match.

      Still, it would be nice if FTDNA and 23andMe handled this similarly.

      I have a dozen matches at 23andMe in the same region. Many are almost exactly the same as yours. Positions 1 to 29.4 MBP. I only have one MRCA for the dozen and maybe it has no connection to the DNA? ALL of these matches have low SNPs. Less than 100 SNPs per cM. That's low.

      You should check the SNPs. You probably have a poor SNP count.

      I have been unsuccessful so far in finding any valid matches on the left side of Chr 15, with the exception of the one dubious match I mentioned above.


      • #4
        I posted a related question on the 23andMe forum and got a couple of very useful replies.

        Chromosomes 13, 14, 15, 21, 22 are acrocrentric. They have very short left arms. Family Finder Chr Browser blocks out these regions and labels them SNP poor regions.

        For Chr 15, the centromere is at 19 MBP.

        23andMe counts the entire left arm even though they dont test it. This is apparent in their Raw Data. So matches are assigned beginning at position 1. That adds an extra 19 MBP, and corresponding cM and gives an inflated value.

        I think FTDNA's approach is better by counting only the region that was actually tested. In your case, 10.57 cM.

        You still have the issue of your 10.57 cM being spread over two pile-up regions. IMHO, a shaky match.

        Keep in mind, that DNA Painter follows 23andMe's lead and allows painting matches in the SNP poor regions.


        Last edited by mabrams; 29 May 2021, 09:50 AM.


        • #5
          mabrams, thanks for the details in your two posts! "Thumbs up"


          • #6
            Really sad to hear that 23andMe is counting an unsampled region! We can see that the unsampled region is about 25 million base pairs, but how long is it in centiMorgans? It's the cM value, the "recombinational map", that is most of interest for genetic genealogy, because it's the recombinational map that reflects the probability of a given segment being transmitted intact to the next generation. The DNA around the centromeres tends to be very repetitive, and as a result it may not contain usable SNP's. Typically, the amount of recombination that occurs within the unsampled "centromeric" region is very small, usually less than 1 cM (we know this from recombination data for markers on either side of the centromere), so we would not be very concerned about allowing a matching segment to cross the centromere, even if there is a gap of millions of base pairs. However, given the bad behavior of this end of chromosome 15, and the apparent absence of a reasonable number of SNP's on the other side of the centromere, I'm afraid we have to discount any matches that seem to extend into this uncharted realm.


            • #7
              Not sure what happened to the color, but the top chromosome has a dark bar on the left. It was purple at 23andMe and on the image I saved.

              23andMe clearly shows the SNP Poor Region on the far left. The 'dark' bar contains 42 cM according to 23andMe. Even in that small bar, there are two pile-ups and a centromere (probably connected to the second pileup).

              Capture chr 15.png

              Its too bad, as that match had an MRCA going back to 1600s Dutch NY.