Sunday, 8 October 2017

Tom Jones on Systematically Using Autosomal DNA Test Results to Help Break Through Genealogical Brick Walls

Widely respected US professional genealogist Tom Jones gave a presentation in a BCG webinar last Friday showing again how DNA evidence has moved into the mainstream.

It's free online through Legacy Webinars until 13 October here. I highly recommend it for serious genealogists - that's despite the comments below.

What struck me was the use of percentages and probability when it came to the DNA evidence. Values from the case study were compared to expectation from theory. Yet despite having quantitative evidence confidence in the agreement was not quantified. And there was no attempt to quantify confidence arising from the conventional documentary evidence.

Also it was not clear there was any consideration that endogamy might have made the relationships seem closer than they are.


Debbie Kennett said...

I thought this presentation left a lot of questions unanswered. First of all there was no attempt to quantify whether or not the shared segments were real. Given that the largest segments were all under 15 cMs, and often with very low SNP counts, there's a high probability that many of these segments would not survive phasing.

Four people triangulated on one segment, and three people triangulated on a second segment. This should have raised alarms bells. There is already a low chance of fifth cousins matching. Given that there are 22 possible chromosomes where you could share a segment, the chances that three or four fifth cousins could match on the same segment by virtue of a single genealogical relationship are very slim indeed. It's far more likely that they are matching through descent from the same population.

The presenter did not provide any information on the family trees of the people who were targeted for testing. With autosomal DNA it is essential to research all branches of the family tree in order to check that there are no cousin marriages that might distort the relationship predictions and also to determine that there is no possibility that the two people could be related through a different pathway other than the one already identified. It may be that this research was done but wasn't included in the presentation.

As you say, endogamy is also a major consideration which was not taken into account. With the small founding population of the US, endogamy is a certainty for anyone with deep roots in the US, and this needs to be factored into the analysis.

The other point that wasn't considered is the time depth of the shared segments. Fifth cousins are in general only likely to match on a single segment, but if you only share a single segment there is a very wide range of possible relationships. As the simulations from Speed and Balding have shown, it is far more likely that segments under 15 cMs are very distant rather than recent:

In conclusion, I think there were many alternative hypotheses that should have been tested and rejected in this particular case before being able to state with confidence that the research question had been answered.

Dan Edwards said...

Debbie makes some good points.

The details of such match analysis is probably beyond what can be covered in a beginner genealogy seminar, though. I often tell people new to genetic genealogy that it is a steep hill to climb, to understand what is going on.

That applies to genealogy too, at least for those who want to tackle ancestors not otherwise obvious in numerous records.