Sunday, December 14, 2014

AncestryDNA Circles Rolling Right Off The Page & Other Oddities

Ancestry and the little genealogists?

I've been working on those novel AncestryDNA Circles, which are currently deep in beta. Why am I  doing this when they aren't currently very interesting? Probably because I'm avoiding doing hated Christmas chores?? Ha, Ha, Ho! Ho!

When I examined my Cousin Nan's Circles I noticed she had at least one that my Mom should be in. Upon further examination we should have 3 more than we have (that we even know about?). We've each had 9 Circles since the Circles started. In an effort to forget Christmas chores I plunged into the Circles. I tried changing the information to match the Circles we should be in. This hasn't resulted in creating any new circles yet? I heard that the Circles cycle every few hours. So I would think I should see any new Circles showing up within a day or so?

I contacted Ancestry via Facebook messaging yesterday. They attempted to fix my Circles problem. Apparently they were not able to add any Circles. What they managed to do is replace one Circle ancestor with another. It appears our Circles are stuck on 9 ( I thought we had 7 but it's 9). Something is wrong which isn't allowing us to gain anymore Circles? This is what I saw last night when they tried to remedy the problem. You can see (below) one ancestor is replaced with "feature not available". This ancestor was later replaced by another ancestor.

Looking at the member trees for one Circle we're looking to join I noticed that one person doesn't even have the ancestor the Circle is based on in their tree. This person hasn't logged in since June? I take it he couldn't have changed the tree since the Circles were created?

After even brief examination it's clear the Circles are deeply mired in Beta. We hope they can fix these problems because previous search function problems remain. I got another message from Ancestry recommending I call them about the problems with our Circles. I learned to call them well before or after lunch. Backup phone Reps. apparently answer around lunch? I believe so because I called about a DNA test kit one time and the person seemed confused. They were like DNA test? Then they remembered, "Oh yay".

So now I have a new Circle but I'm still missing 3 Circles we should be in? If we can get them all at once I would be satisfied. The Christmas chores are looking better now.

Wednesday, December 10, 2014

DNA: Grading Autosomal Testing Companies

What I've been doing so far with my test results is supporting my paper trail trees with DNA evidence. I haven't broken down any brickwalls yet, but I'm finding some evidence which may do that in the future.

Having worked with the Autosomal DNA tests since 2012 I've formed some opinions about the different companies I've had experience using.
Here are my grades based on my experience:

Click to enlarge

Since the point of testing for me is to confirm relationships my overall scores reflect the tools provided by the companies to do this. Some explanation of my scores:

Family Tree DNA: B Overall. Because they have good tools. If they used phasing they would get an A- overall. They need to improve their family tree interface and ethnicity predictions.
AncestryDNA: C Overall. They would get an A if they had tools for confirmation of relationships. The trees here can be so horrible which creates a problem. Some are great, but others just make me crazy. They do have a new tool called Circles which is mainly a novelty. They provide good matches; which, unfortunately need confirming elsewhere.
23andMe: C Overall. This grade reflects the low query response rate, and the low level of family knowledge of the database matches.

I looked at this review of AncestryDNA yesterday and it inspired me to grade the companies: My AncestryDNA Review: A Cautionary Tale. Interesting post.
The International Society of Genetic Genealogy has a great page comparing the DNA testing companies.

Sunday, December 7, 2014

Are The New Moderate Matches Good?

I had to correct this post again. Timber is a filter that AncestryDNA uses and could be a reason for the segment size differences? Breaking up of compound segments through phasing is another possible reason You can view an explanation of Timber here.

Some people feel the Timber system is flawed? I would like to know more about it.

Because of the possibility of a phasing error I wouldn't completely disregard moderate matches. Some of these share large segments of DNA not reflected in the phased results. There is a phasing error rate of 1 for every 100 heterozygous sites in a sample.

Since GEDmatch is now working again I decided to do some research on my AncestryDNA matches. I was curious about those Moderate about 6cM and under matches. I did some searching on my Moderate match usernames to see if I could find any to compare with. A few had uploaded to GEDmatch and I quickly discovered the cM ranges they gave for Moderate matches were way off. I found that one of these matches shared a 15 cM segment and another an 18 cM segment, while yet another shared an 8 cM segment. Why put out these estimates at all? Why not explain these are Timber filtered and phased so the results can't be compared elsewhere?

The illustrations below demonstrate the differences between the Timber filtered and phased results from AncestryDNA and the unphased, unfiltered GEDmatch and Family Tree DNA results.

The Extremely High match I found is obviously extremely good. Good call here.

Now we move to a Very High Comparison. We share a couple segments so I'd agree.
Just plain High. Ok we just have on segment so I'll go with that. It's near the range.

Now we move to Good matches which are sometimes better than High? I noticed that these Good matches also match me at Family Tree DNA and the segment sizes are basically the same there, and they declare these matches to be 2nd to 4th Cousins instead of distant.

Good Confidence match

Good Confidence match
Good confidence match
Moderate matches are all over the place cM wise. Likely range from 5 cM's to 20 cM's.
Moderate Confidence
15 cM segment Moderate Match

Wednesday, December 3, 2014

How DNA Segment Data Has Helped Me

Big time triangulation

When I first took the Family Tree DNA, Family Finder autosomal DNA test, I had no idea how I could use the segment information to prove my matches. It seemed very complicated. I quickly found out the mechanics of it weren't that complicated at all, and it's necessary to prove relationships with cousins.

The first cousin I confirmed through DNA was a 5th cousin I had been collaborating with for several years. We could be fairly certain our segments came from our paper trail lines because we don't appear to have any other relationship according to our trees. When my Mom tested and shared the same segments we knew for certain they were IBD through my maternal line. We also could be sure we were not related through my mother's maternal side which is Nicaraguan, and my cousin has no Nicaraguan ancestry.

When examining the segment information for my 5th cousin we notice she shares more than one segment. I have found if someone shares more than one segment, including one over 10 cM's, and the others about 5 cM's or over, a relationship within the past 6 generations generally can be found. Also my 5th cousin has had her brother and aunt tested adding confidence to the IBD segments. We can see her brother shares a 5 cM segment which is IBD because 4 of us share on the same segment, and the others segments are in the high confidence, IBD size, range. All of us sharing these particular segments are descended from double first cousins. So we cannot say from which ancestral couple we received these segments? If we can find other cousins who match on these segments and they are not in our double first cousin line we could then say we got these segments from a particular ancestral couple.

I've been able to identify how other cousin matches are related in a similar way by fishing for them in my matches. I was only aware of one cousin in the Family Tree Data base when I started. By looking at the trees of my matches at Family Tree DNA I was able to find possible connections. When I spotted a shared ancestor I would use the "in common with" feature provided by FTDNA to see others who share DNA on the same segments. I would then check to see if we had triangulation by using the chromosome browser tool. By comparing these matches 5 at a time I can spot individuals who share DNA on the same segment. If this person, sharing on an identified segment, had no tree posted I would email them to let them know we may share a common ancestral line based on segment sharing. I've been able to identify additional cousins sharing on the same segments that way.
I also check for ICW matches when I get a new match and the cousin relationship is closer than remote cousin.

Attempting to identify every possible match helps to firm up the segment data, and cross off some matches from the list. This could help brickwall related matches to surface by reducing the number of matches.

Matrix showing nearly
everyone matching
Through the process I described I was able spot 15 people sharing DNA in the same place on Chromsome 18. All of the shares are in the presumed IBD cM range. This could be a pile up area, but one person also matches with me on another chromosome, suggesting a closer ancestral connection rather than a pile up. Most of these matches don't have trees so it's impossible to draw a conclusion on who the common ancestors are?   It appears that the 2nd to 4th cousin match sharing a 33 cM, and 18 cM, segment might be related through the Wray family? The Wrays migrated to Allen County, Kansas in the 1850's. This particular match also had family in Allen County, Kansas. The surname Ray appears in another common match's tree. The places listed on a few other ICW matches trees also point to Wray/Morris being our shared line. I can't say all 15 people are from this line without more proof. I was reading Roberta Estes' blog the other day and was reminded about the Matrix which is another tool at Family Tree DNA. With this tool I was able to learn that all of the matches on segment 18 also match my 2nd to 4th cousin, with the Allen County, KS roots. This is encouraging me to pursue this further.

Triangulation of matches
from the same area in Ireland
I have not been able to give my Irish and Nicaraguan segments family names. The only family surnames I know of in Nicaragua are Del Castillo and Garcia. In Ireland the only family surnames I know of are Mullen and Huvane/Huane. I have one Del Castillo match. I haven't found any cousin matches based on the other names. I always keep an eye out for these names.

The Nicaraguan matches don't usually have exact locations in Nicaragua. I know my family is from Granada, Nicaragua, so I keep an eye out for those matches. My Grandmother Graciela Del Castillo-Forgey had a number of Nicaraguan friends. I am familiar with their surnames, and I'm finding those names in our matches, names such as Lacayo and Chamorro. According to my Grandmother first cousins often married in Nicaragua so it may be difficult to separate out these DNA segments due to so much intermarriage. This may also be true on my Irish side, because the population was small in our ancestral locations. My family came from the Townland of Pollaturick in Galway, Ireland, and the Townland of Kilvine in Mayo, Ireland. I find it rewarding when my matches ancestors come from the same small area of Ireland as my family. I have a triangulation with several matches from the same small area of Ireland. Since most genealogical relevant records for Ireland have been destroyed I don't expect we'll ever find the names of the ancestors who passed these segments down to us. It's great to have the ancestral locations confirmed through DNA anyway.
Itty bitty Owens segment. An unnamed company would call this a moderate match??

When it comes to brickwall lines and segments I haven't been able to use the segment data, yet, to absolutely prove a relationship. We need more segment matches to add weight to our paper trail inferences. Right now I only have segment data for single segment in a particularly challenging brickwall line. I have brickwall related matches at fill-in-the-blank -------- Company. They don't provide segment data. So I have no segment data on for these matches, I won't be able to triangulate or confirm these matches. The segment I have is only 5 cM's, which I can't be confident is IBD. It would be great to find more and larger segments, and triangulation on the brickwall lines.

How segment data has helped me:
  1. They've provided actual proof that I'm related to a cousin. Without having the segment information there is no proof you are related to anyone. If a company provides you with a list of match names it's meaningless. We need to cite the DNA evidence like we cite documentation on our tree. We need to cite numbers and chromosomes. Not having this information is like only looking at an index to records without seeing the actual records. You don't have possession of the evidence without segment data. 
  2. I've been able to infer, then prove, a relationship with several matches based on matching on the same segments. Some of these people didn't have trees. When I emailed them they would often confirm the relationship through a particular line.
  3. The segment information allows me to evaluate which matches are the most promising for further study; such as those sharing more than one segment.
Problems with segment comparison
  1. Segment comparison in an endogamous population is not useful at this time. It only works well in mixed populations where there are more differences in the DNA
  2. Many of us with Early American ancestry find a web of relationships among our matches which is hard to untangle. If you test enough family members this can often be over come.
  3. Segments under 5 cM tend to be IBS and should not be used to draw conclusions. Some people are unaware of what the differences in segment size mean, and draw conclusions using IBS segments.

I've been able to work the DNA results by dividing them into piles based on ancestral locations. My family has such a varied ethnic heritage, making this possible. I can also divide matches by my maternal and paternal sides, because my Mom tested. My Mom's line can be divided in half because her mother was Nicaraguan, and her father was Northern European American. If I find a Nicaraguan match I can be confident it's from my Del Castillo/Garcia line. When I name segments I assign them to a couple based on the earliest shared ancestral couple on our trees.

The segment information has been the most useful, to date, in supporting my paper trails. Sometimes my paper trail is confirmed through shared ancestral names, other times it's shared places.

I have to say seeing the shared segments is rewarding. I mean actually seeing the shared segments with my own eyes. I think it's more exciting than seeing a list of match names.

Some blog posts about segments from authorities on the subject:
  1. "The Folly of Using Small Segments as Proof in Genealogical Research" Part One. By CeCe Moore
  2. "Small Matching Segments – Friend or Foe?" By Blaine Bettinger
  3. "Chromosome Browser War" By Roberta Estes