The Case of the Missing Matches

Soon many DNA matches will quietly vanish from our Ancestry DNA match lists. Will you miss any of your matches?

• I will miss Elaine, a 90-something woman who has amazing knowledge about the complex web of our German roots.
• I will miss Mandy, a young mother who may connect with either German or Swedish ancestry, but is a fun contact.
• I will miss Lou, an African-American man who is probably a descendant of my Alexander ancestors, slave owners in South Carolina.

I told you about Lou a year ago in The Case of the Missing DNA. Comparing notes from various DNA tests and web sites, Lou and I discovered that Ancestry had ignored about 10 cM of our matching DNA, placing our match at a low 8 cM.

Now Ancestry plans to drop matches that are under 8 cM. Will I keep Lou as a match? What matches will Lou miss? Will he be able to solve the mystery of his South Carolina roots?

It's a challenge for the descendants of enslaved persons to find ancestors before 1870. Loss of low-level matches will increase the difficulty of finding those distant cousins. In this time of racial enlightenment, Ancestry is moving in the wrong direction.

Instead of hiding matches and ignoring strands of DNA, Ancestry needs to improve their data science capabilities. Yes, as a data scientist, I understand that is easy to say and hard to do. It's time for Ancestry to do the hard work.

DNA and The Butterfly Effect

Would law enforcement be able to use genetic databases like GedMatch to solve crimes if Ancestry had provided a chromosome browser to their customers?

The CEO of Ancestry this week cautioned that consumers need to be careful about the companies with whom we share our DNA. She pointed out that Ancestry has the highest standards around privacy and does not cooperate with law enforcement.

I almost fell off my chair laughing. In another post I'll explain how Ancestry reduced privacy options recently in a way that has me highly concerned and caused me to unlink several DNA tests from my tree.

The CEO's statement made me think also about cause and effect.

If Ancestry had provided a chromosome browser, they would have been able to serve their customers who want hard science in addition to warm fuzzy feelings. We would not have needed to turn to GedMatch and other companies for this service.

Additionally, if Ancestry had accepted DNA uploads from other companies, they could have provided cross-company matching and captured a larger share of the market. Instead they ceded this market to GedMatch and now to other companies.

I did a (non-scientific) check of my top 3000 matches on GedMatch. A full 50 percent of my matches were from Ancestry. And my family and I have better privacy on GedMatch than on Ancestry, which is a story for another day.

My opinion is that Ancestry's choices have been an important cause of the growth of GedMatch. I'm personally on the fence about law enforcement using genetic databases. However, I also think Ancestry needs to own its part in the situation, rather than denigrate companies that fill gaps in Ancestry's DNA offering.

So I ponder if the Golden State Killer was caught thanks to a long-ago decision made by Ancestry. Was it cause and effect?

The Case of the Missing DNA

A recent inquiry from a DNA match sent me into the mysteries of AncestryDNA matching. I was appalled at what I learned.

In the field of data science, we use the term "single source of truth." You would expect that DNA matching would be consistent across all the different testing companies, so any and all of them would be a valid source of truth. Not so. If you skip the story, please take a moment to look below at the explanatory graphic.

It started with a routine email from a gentleman I'll call Lou. He found our match on Family Tree DNA and asked if it was possible we were related through a particular surname. He closed by telling me he was of African-American descent.

Knowing the challenges faced by African-Americans who are researching their roots, I gave the email far more attention than I would have if it had been from a Caucasian match.

Lou and I had no shared matches on FTDNA, so  I turned to Ancestry, where I have identified many cousins from that branch of my family. But there was a problem. Our match on Ancestry was only 8 cM (1 segment), where on FTDNA it was 27 cM (3 segments). How is that even possible?

Our FTDNA match includes a couple of short segments. Eliminating them leaves our FTDNA match at 19.9 cM (1 segment), still more than double what AncestryDNA showed. We exchanged some emails to discuss the source of the data.

• Lou had uploaded his AncestryDNA results to FTDNA, MyHeritage and GedMatch. 
• Lou had tested with 23AndMe and uploaded that result to GedMatch.
• I had directly tested with FTDNA and MyHeritage, in addition to AncestryDNA. 
• I had uploaded my AncestryDNA result to GedMatch. 

All match combinations except AncestryDNA are in the range 17.8-19.9 cM, with one segment in the same approximate range in chromosome 12. Of course we can't see what AncestryDNA is suggesting.

Choose Your Source of Truth

My brother's match to Lou is also in a similar range to these numbers. At Ancestry DNA, Lou's brothers and daughter have a stronger match to me than Lou does. However, none of them have uploaded to GedMatch, so we can't see the science behind the numbers. This particular chromosome range does appear to fall in or near an ISOGG-documented slight pile-up area.

So this leaves the possibility that AncestryDNA has chosen to ignore some of the match due to pile-up. Wouldn't it be nice to be told that?

Does Ancestry have a computational error on my match with Lou, since Lou's daughter matches me more strongly than Lou matches me? Or does she match me in an entirely different way via her mother's lines?

What is the best source of truth? If you are using only AncestryDNA for your DNA matching, you are not seeing the whole truth. GedMatch is free. FTDNA is inexpensive. MyHeritage has some great tools. You can choose your source of truth.