Tuesday, May 10, 2016

To Phase Or Not To Phase? Plus, She's Back?

My to Quack or not to Quack souvenir  from my recent vacation

 

That is the question...

Is the phasing and filtering AncestryDNA does worth the extra processing? I've wavered about this for years. It sounds like a great idea. On a theoretical basis it is. In practice not so much. The phasing AncestryDNA does attempts to use haplotypes to separate the DNA we inherit from each parent. The results are also filtered in order to remove matches who share population segments. When I listened to an Ancestry representative explain the haplotype method she did say there was an error rate with the phasing. Some haplotypes haven't been encountered before. Removing population segments, with filtering, isn't helpful for me since these segments at least tell me which ethnic group a segment comes from.

A recent update to the AncestryDNA product has many discussing the merits of this companies approach to matching. An issue was brought up at the Facebook ISOGG group which I hadn't noticed. Before the recent AncestryDNA update parents and children were said to share up to 90 segments of DNA. According to the other companies around 30 segments are shared. This vast discrepancy is due to the fact Ancestry's phasing and filtering chops up segments. The recent update has brought the number of shared segments down to the 50's. Still many more than the other companies.

I've posted about the fact that during AncestryDNA's previous update, when Timber was introduced, a 3rd cousin went missing. I checked to see if she was returned immediately after the update finished. I didn't see her listed as a 3rd or 4th cousin so I assumed she wasn't returned with this update. Going through hints yesterday I found her. She has returned. She is listed as a 5th to 8th cousin now. What does this tell me? First of all the prediction is off. Secondly it tells me the AncestryDNA product is still in a state of flux and who knows what will happen with future updates? Apparently someone can match you today, and may be removed in the future, just to possibly be returned somewhere down the line?

I didn't find that any close matches were removed this time. I did find several distant cousin matches had been removed. It's possible these matches do match DNA wise? AncestryDNA states it is possible good matches were removed.

I would rather see AncestryDNA do away with the phasing and filtering. Shuffling matches in and out of our lists doesn't make any sense. It's just confusing. Does the phasing and filtering improve match results? Not in my case. Predictions at the 3rd cousin range and beyond are impossible to get exactly right. I'm not even sure if phasing and filtering helps improve predictions for closer cousins? What it can do is remove good matches.

One thing is certain, every time AncestryDNA updates results they get publicity. As someone once said "there's no such thing as bad publicity."

2 comments:

Kalani said...

Even before Timber, and now this recent change, all my predictions aren't all perfect either. I didn't take a good look at all of the differences yet, but they do vary, especially with the more distant matches. Then there is the issue of half-relationships and throw in endogamy, not everyone will get the predicted matches right. This is why it is important to compare trees, work on collateral branches and their descendants, and continue to work as far back as possible.
https://www.dropbox.com/s/iox41yu5bi5nbjd/COMPARISONS-sorted.pdf?dl=0

Annette said...

True Kalani. Good trees are the most important factor when it comes to making connections with matches.