Back in July, Ancestry started to roll out matches by cluster. Already familiar to users of Genetic Affairs, DNAGedcom, MyHeritage and Gedmatch, it has been great to see this capability built into AncestryDNA and available to anyone with their Pro Tools plan.
Last month saw an important enhancement: the long-awaited AncestryDNA custom clusters. This adds two key features:
- You can specify a custom cM range (down to 20cM); previously this was fixed at 65-1,300cM
- You can also now focus the clusters on a specific match in contrast to the standard clusters, which are based on just the tester
Rather than dive straight in, I was keen to try this feature out in an informed way with the help of Angie Bush’s introduction and Rick T. Wilson’s recent article with case studies. These articles helped me to understand the concepts and steps. Overall, it’s a straightforward process and I think Ancestry have done a good job with the user interface. Inevitably, some of the jargon can be slightly confusing at first, so I found it useful to read through the articles and the examples in them.
Creating a custom cluster
To make a custom cluster, visit your matches page at Ancestry DNA and then click By Cluster. After you click on Create Custom Cluster:
- You’re asked to choose a match of interest
- This essentially means ‘someone to use as a basis for the clusters’: they will consist of people who match both the tester and this person
- The instructions suggest you choose someone who is related to you via a ‘brick wall’ ancestor
- The intended goal is to focus on those matches who might help you break through brick walls
- You can then choose up to four other matches, which Angie Bush refers to as sidekick matches. These are optional; if you include them, they should be other matches who are
- Via the same ancestor as the match of interest
- Share an amount of DNA that exceeds the centimorgan range you choose
- Finally, you’re asked to specify this centimorgan range
Having understood this, I was ready to try it out on a couple of brick walls in my direct line.
My brick walls
1. Mystery Moore parents in Havant
First, I returned to the Moore mystery I examined recently:
- Who were the parents of James Moore born c. 1786 in Havant, Hampshire?
As I explain in the previous post, due to location, naming patterns and professions, I believe James’s grandparents were Edward Moore and Mary Atwick (born about 1723).
I know that trying to solve this puzzle with autosomal DNA is a bit ambitious, since the unknown ancestor couple are a set of fourth-great grandparents. This is sufficiently distant that the amount of DNA shared with relevant testers is likely to be small. The biggest confirmed match, NH, is a descendant of James and Elizabeth’s daughter Martha.
Since I’ve tested four people in my father’s generation, I was able to choose his cousin VB, who shares the most with NH.
An important reminder: if you can, definitely test as many oldest generation people possible while you can! However, VB and NH, who are fourth cousins, still share just 24cM, so that doesn’t give me a lot to work with.
After clicking add custom cluster, I entered NH as the match of interest. I didn’t have any sidekick matches so I skipped this and entered a range of 20 to 60cM. My rationale here:
- Anyone sharing more than 60cM will be a known cousin in the 3rd cousin range
- Therefore, if I included them, the results would be polluted with other matches not relevant to the Moore line
I hoped I might see two distinct groups of matches, one consisting of relatives of James’s mother and the other of relatives of his father.
In fact, I got just two small clusters, with three people in first, and the same three people plus another in the second. This may be down to the small testing pool for English matches, or bad luck, or just because my match of interest was so small.
NH’s daughter featured in both clusters, leaving just two matches that I hadn’t previously seen: SF (25cM) and MP (24cM). Contrary to my expectations I was actually able to place both of these people in my tree, perhaps because of all the work I did researching the previous article:
- SF turned out to be a descendant of Edward Moore and Mary Atwick via their daughter Susanna
- MP is a descendant of James Moore and Elizabeth Jones via their daughter Martha
I then looked at my dad’s cousin VB’s shared matches with both SF and MP. At this point Ancestry’s SideView feature was helpful, highlighting some tiny matches that were actually on VB’s maternal side rather than the relevant one. There’s one match with a reasonably significant looking match to SF. Although her tree was private, searching shared matches revealed that it seems to contain both the names Moore and Palmer. So I found myself in the bizarre situation of messaging a user based on a 9cM match!
So while this didn’t give me a huge breakthrough, it did prompt me to identify two new matches on this line, and has also given me my first solid genetic link to Edward Moore b. 1723.
2. My Marshall line from County Tyrone
Feeling that my Moore mystery is possibly too distant to offer up any quick insights, I switched my attention to a closer mystery on my mother’s maternal side: the identity of the wife of her 2nd-great-grandfather Joseph Marshall.
Again, my mother’s cousin (RM) has better matches on this side than my mother:
- Two strong matches, JL (118cM) and KR (102cM) are descendants of Mary Ann Marshall born circa 1819
- Interestingly, JL and KR share only 119cM with each other
- I was confused by this but realized their relationship is half second cousin. 119cM is a typical amount for this relationship, as illustrated in the shared cM tool
Mary Ann appears to have been a sister to my mother’s great-grandfather Matthew Marshall. What do I mean by ‘appears’?
- Sister to Matthew is the closest that I can speculate that they could fit in based on known genealogy
- Based on this, RM would be 3rd cousin once removed to JL and KR
- 118 and 102cM are very high amounts to share with a 3C1R but are in range
- Unless there are additional relationships (quite possible!), these matches can’t be any more distant to RM, so for now my hypothesis is that this is the connection
I can also check how much my mother and her cousin share with JL and KR:
- My mother shares nothing with either of them (again, this is within range for a 3C1R)
- Our other tested Marshall cousin shares nothing with JL and 18cM with KR.
Ancestry’s sideview assures me that this is a maternal match for my mother’s cousin (as opposed to Both or Unassigned). So my current view is that we’re looking at some Marshall DNA that my mother happened not to inherit but that her cousin RM did.
This time I used JL as my match of interest but added KR as a sidekick match. I used a range of 20 to 100cM. The outcome was similar to my first attempt: just a few small clusters, with four core people appearing in all of them.
Within these clusters, two matches lead back to unknown Marshall people, but typically for Irish matches, I can only take them back to the 1850s. If anyone reading this has a lead on Mary E Marshall b. 1855 County Tyrone who married William E Scruggs and died 7 Sep 1916 in Georgia, let me know!
Looking forward
I’m sure there’s a lot more I will be able to do with these when I have more time. There are also some nice features, such as the ability to add a coloured dot to all cluster members at one time.
Overall, as someone who
- doesn’t have quite the volume of significant matches that some people do, and
- is already obsessive about examining shared matches
… this feature hasn’t added an enormous amount of value in these first experiments. However, I look forward to trying it out some more. In the meantime, I’m grateful that it helped me to focus on these specific lines for a sustained amount of time! If anyone reading this spots something obvious I’m missing or has any suggestions, please feel free to get in touch. Thanks!
Contact info: @dnapainter.bsky.social / jonny@dnapainter.com