DNA match results from 23andMe, AncestryDNA or Family Finder tests provide a one-dimensional view of your relatives. Each service reports a list of individuals where identical segments of matching SNP data indicate recent common ancestry (i.e., Identity By Descent or IBD matching). The analytical methods used to identify these DNA relatives are not perfect, thus these lists are plagued by two types of errors: 1) Individuals are identified as relatives where the underlying genetic identity was inherited from a common ancestor(s) that is more distant than the most recent common ancestor(s) you share with that matching individual. 2) There are also individuals in the database that should be included in your list of relatives, but are missing due to the criteria used in the analysis. The computational algorithms each service uses to identify relatives are conflicted between these two types of errors, thus methods and settings that decrease one type of error increase the other type. Elizabeth Thompson from the University of Washington has written a comprehensive review of the theoretical and practical considerations of this identity by descent between relatives.
Comparison of genetic matches in common among multiple individuals is a method to increase confidence in genetic identity due to recent common ancestry. Essentially, if individual A, B & C are each descendants of a common ancestor, B & C may be matches to A, while A & C may be matches to B, and A & B may be matches to C. Detection of these common matches is the basis of DNA Circle formation by AncestryDNA. Recognize that I said that descendants "may be" matches. If a pair of common ancestors is beyond four generations removed from each descendant, it becomes increasingly likely that two descendants didn't inherit a common segment of DNA from these ancestors. These are not the errors of omission mentioned above, these are cases where due to the 50:50 segregation of chromosome pairs each generation no region of the autosomal genome is shared between the descendant individuals. Confirmation of matching among more descendants connected through the common ancestor(s) of interest, therefore, increases the genetic support for common ancestry. The DNA Circle function in AncestryDNA automates this process and produces nice images of these comparisons. Unfortunately, the family tree information needed to form a DNA Circle within AncestryDNA is restrictive, thus precluding circle formation in branches of a family tree that end with a brick wall. This is the case for my McAllister family line.
Eight descendants of our Irish immigrant ancestors shown in the pedigree below have tested with AncestryDNA. From historical documents, the immigrant family includes mother Mary and her sons Patrick and Thomas McAllister. Because we don't have information on Mary's parents, AncestryDNA will not form a DNA Circle for her. Furthermore, we don't know the identity of Patrick and Thomas' father, so they also will not form DNA Circles. I have communicated with DNA matches in this family to try and gain information on whether matches with my dad are common matches with them. I have also requested that matches share their results with me. Both approaches have not been very productive. Thus, I enthusiastically welcomed the addition of the Shared Matches feature in AncestryDNA. This is analogous to the "Matches in Common" feature of the Family Finder test of Family Tree DNA; however, only a single member of this family besides my dad is in their database. Although this is a welcome addition to AncestryDNA, it is also important to recognize its limits due to an imposed 4th cousin cutoff.
Here I summarize the primary match results, and the shared match results for the three tests (A, B & E) that I have access to. On the top row for each test's results, the primary matches are identified with the relationship estimated by AncestryDNA (close family, 2nd cousin, etc). Note that in the DNA matches of individual A, individual H is missing. In contrast, individual H is present and individual E is missing in the matches of test B. These are potential matches across the deepest point of common ancestry, and given that each lineage is at least four generations removed from the common ancestors, the absence of a matching DNA segment or the failure to detect one if its there is not surprising. This is especially apparent in the primary test results of individual E, a descendant of Thomas and five generations removed from the deepest point of common ancestry. Looking for matches to Patrick's descendants, only individual A is identified as a match.
The Shared Matches feature of AncestryDNA, allows the user when viewing a primary match to also view shared matches that are 4th cousins or closer to the tested individual. On each of the rows below the primary match relationship estimates, a primary match is used to identify shared matches. For example, looking at test results for A, individuals C & D appear as shared matches with individual B. Although F & G are listed as distant relatives in the primary match results for individual B, they are missing as shared matches with B in the test results of A (see Figure). In contrast, A appears as shared match of F & G in the test results of B, because each of the secondary relationships (e.g., F-A & A-B; see Figure) is below the 4th cousin cutoff. The same pattern appears in the results of individual E and match G.
Beyond the contrasting pattern between 4th cousin estimates and distant relative estimates, what these comparisons illustrate is that sharing the primary results is still the best way to explore common matches within AncestryDNA. Lines in the pedigree above are weighted by the DNA matches supporting each connection. The fact that the test results of A has E, F & G as primary matches, and that the test results of B has F, G & H as primary matches strengthens the support for the sibling relationship between Patrick and Thomas. The relationships are further supported among the descendants of Thomas' children with the test results of E showing G & H as primary matches. Because of the distance of the estimated relationships, these are not reported as shared matches. Much of the open space in the table above is also likely not being reported as shared matches due to the 4th cousin limit. Shared test results empower the AncestryDNA platform for those users that want to interrogate the branches of their family tree, and it would be great if more users were willing to use this feature. I'm grateful to those relatives that have shared their results with me.