Archives

The Reinvented 23andme

Jun20

My cousin Corinne tantalized me by showing me some of the interesting new features at 23andme, so I bought the currently discounted upgrade and soon sent in a new vial of spit. As the email I got pointed out, my test used an older chip and more information is available on the newer one. Father’s Day is the last day for the current special.

Next I opted into the subscription for Premium Plus membership which provides clustering,  historical matches, reconstructed ancestors, and many more matches as well as updated haplogroup information. One of the main things I wanted was the ability to get the exact segments where my relatives match me in a chromosome browser; a feature that had been turned off after the break-ins a while back. You cannot download all your segment data but you can view up to 5 relatives compared to yourself or someone else you match. You can get to that comparison by clicking on the blue “Compare with more relatives” at the bottom of the panel which shows your DNA chromosomes with shared DNA. Then at the bottom of the comparison page you can get the exact numbers.

The top of an example comparison page, my dad to multiple relatives. Notice that it indicates that he shares one fully identical segment with his nephew PG (on the X as shown below)

I have written many blog posts about using segment data (click here). I maintain a spreadsheet for all my dad’s segment matches and note where my brother or I have the same match, as well as other known relatives. Often I can tell what ancestral line a new match is on from their matching segment information. I have alot of updating to do now that this feature is back!

At the bottom of the images of chromosomes and segments there is the numerical data, which can be cut and pasted into a spreadsheet

You can also get to the attractive new clustering page from the relatives in common section of the match page as shown below. I have many blog posts about clustering (click here). An advantage of the 23andme implementation is that you can adjust the parameters. Probably I will need to play with it more and give it its own blog post.

Example of relatives in common buttons to cousin Elizabeth

One of the other new features that I was interested in was the comparison to ancient DNA, in other words to ancient bodies whose DNA has been sequenced.  I was rewarded by discovering that I share a piece of DNA with Otzi the Iceman! Small but exciting.

Continue reading →

Further experiments with AI and genealogical documents

May23

A recent 2nd cousin DNA match (click here for that story) inspired me to explore German records from Bavaria. So I decided to try AI to read the handwriting and translate it. First however, I did some more work on my Norwegians.

My home page at Claude

There is more than one free AI out there, so I ran an experiment. First I submitted a not great scan to all three of the better known ones, Claude, ChatGPT and Gemini. This is what I found:

  • Claude told me it was hard to read and only gave me a partial.
  • ChatGPT made stuff up that sounded totally plausible, but did not seem to be anything like the original Norwegian.
  • Gemini made an effort to translate it and did well, but organized it into bullet items.

So my new method is to upload the same document to Claude and Gemini and merge the result. One thing I like about both of them is that they remember my recent efforts in the left hand sidebars (on Gemini ask for the sidebar in settings). So far I have not used either one heavily enough to be asked to upgrade to a paid version, unlike ChatGPT which has already requested that.

Next I tried some handwritten German documents. In my opinion Gemini did the best job. Read on for how each AI performed.

Continue reading →

My Fastest Father Find Ever

Jan9

Randi contacted me for help finding her unknown biological father. I advised her to test at Ancestry and get back in touch when her results were ready.

When they came in, it took only three hours to find Randi’s unknown biological dad! She had two second cousin level matches at Ancestry with good trees who did not match each other. That meant that the search would likely be easy, since all I had to do was find where those two trees intersected. Here is how I did that, step by step.

First I created a private searchable tree at Ancestry to use for this case. I started it with Randi and her mom. planning to make floating branches for related people by copying the relevant lines over from their trees.

The best DNA match on Randi’s unknown father’s side was Brad at 322 cM. So using ProTools, I sorted the matches Randi and he shared by his closest matches, as shown in the image below.

Clicking on the sort button brings up a box where you can select to sort by the match’s relationship

The idea was to find the common ancestor among those matches. This would be the line that Randi is related on. To do that, I should have looked at the best one with a tree, excluding close family, but I saw that there was one a bit further down the list, Bob, who had an unusual surname, call it Roper, that was the same as one of Brad’s great grandmothers. So I built her tree back up a few generations and down again. I then copied Bob’s paternal ancestors over, looking for an intersection. I did not find one, so I moved on. Later Brad told me that the error was two women with the same name and birth year incorrectly in various trees, including his.

So I went back to the common match list and found the best match to Brad with a tree (Peggy). One of her grandmothers shared a surname, call it Whistler, with the husband of the Roper great grandmother. So I built the Whistler tree. Quickly found a common ancestor for Peggy and Brad with an unusual first name born in 1830. Built the tree of all his descendants. Somewhere in that tree will be our man.

Time to look at the other possible second cousin match, Jim, at 268 cM. The plan was to repeat the process of finding the common ancestor to his best match with a tree. However, his surname, call it Wander, had already showed up in the Whistler tree. Having collected all the descendants of Brad’s Whistler great grandparents, I noticed that one of them had married a Wander. Was that Wander in Jim’s tree? Yes, she was his aunt!

That Whistler-Wander couple, who must be Randi’s ancestors, had two children, a boy and a girl. The boy was the right age and in the right location. Could it really be this easy? Yes. The details of his life fit what was known. Since Randi’s presumed father is Jim’s first cousin, Jim is Randi’s first cousin once removed.

Continue reading →

Ancestry’s Updated Ancestral Origins

Oct13

Ancestry.com has just completed a major update to its DNA bio-ancestry predictions that has broken down European countries into many smaller areas. They have added 68 locales, mainly in Europe. French Canadians are no longer listed as France, now they are either Quebec or Acadia. They have also made four jewish groups from one: Central and Eastern Ashkenazi plus North African and East Mediterranean Sephardic. Almost all my Jewish friends have no Sephardic listed, giving the lie to some family stories. There are some interesting new categories like Germans in Russia and one combining both sides of the English Channel into a new category called Southeastern England & Northwestern Europe.

I am enjoying looking at the updated origins of many of the people I have helped. On the next page I will show images of the more interesting mixtures. Here is a classic North European mix of a Tenneseean with colonial roots.

I do recommmend scrolling down and clicking on the link to what has changed. Below is the image of how the predictions for our Tennesseean are different from last year. Notice that Germany has gone from 21% to 2%. Some of that must now be Denmark and some SE England & NW Europe.

While 1-2% could be a 4th or 5th grandparent, it could also be just noise or too far back to find. I have often told people not to worry about a 1% call as it is too small to be sure of. In my own case, I lost my 1% Finnish, which I actually thought was accurate because its location matches one of the two where my Norwegian American dad has Finnish, according to 23andme. Additionally, 23andme has my brother matching the other location. We have yet to find that Finnish 4th or further back grandparent, but we do match many Finnish people on those segments over at GEDmatch. Perhaps they just have some Norwegian. More on that in another blog post.

To quote my own blog post from 2020 comparing the origin predictions at each company “predicting the ethnicity for people of Northern European heritage like my brother and myself, is very hard to do accurately because there was so much mixing of those populations.” The book “Who We Are and How We Got Here” by geneticist David Reich makes that point well and goes into the details of what ancient DNA teaches us about European migrations and mixtures.

I found the explanatory Ancestry white paper (click here to read it) hard to understand, so I enlisted two different AI platforms to summarize it for me. A fun use of this technology. To be fair, after the quite technical many paragraph summary ChatGPT gave me, I wised up and asked Claude to summarize it in plain English. That went better. Here are a few take aways:

Continue reading →

23andMe Update

Sep25

23andMe is now owned by​ Anne Wojcicki’s new non-profit The 23andMe Research Institute. An email has gone out to all of us users (see next page) explaining this, which also makes it clear that the focus will be on discovering the health issues in our DNA. Click here for my previous post on 23andme explaining why ​Anne has this interest. 

While crowd sourcing DNA health issues is certainly a worthy cause, it is not the main interest for many of us. Most of my readers consider genetic genealogy a tool to find relatives and uncover family stories.

How the former chromosome browser at 23andme looked comparing a new match to me and my family

The feature I miss the most at 23andme is the chromosome browser that showed ​where my genome matched that of my relatives. Because I have collected an excel file of all my father’s matches with notations as to which I or my brother share, I could frequently tell which family line a new match belonged to,  even a one segment match with no clear relatives in common. ​ Click here for the blog post where I discuss this.

I no longer recommend testing at 23andme unless you want health results or are an adoptee and thus need to test everywhere.

The next page has the text of the email they have sent all of us users.

Continue reading →