Low Confidence Matches
Many knowledgeable users at Ancestry.DNA continue to find the low confidence matches to be a wealth of solid cousins. So much so that sometimes most of your hints lie in those deep dark depths at the back of your DNA matches. Up until now, I haven’t put much thought into it just assuming it was the randomness of DNA. But today, I woke up with a better answer and likely the main reason why those small matches need to be fished through for good family hits.
Example of Low Confidence Matches
Recently, in a discussion thread on 23andMe, someone asked again, if they needed to look through those low confidence matches and I supplied this information:
On the tree I’m currently working on, I have matching hints at:
1 4-6th cousin at the 7th cousin level. This person matches me on other lines closer (French Canadian)
6 Moderate 5-8th cousin level (2 at 4th cousin – remainder at 7-8th)
6 Low 5-8th cousin level (7-8th)
9 Very Low 5-8th cousin level (7-8th)
Another user contributed:
I have 15 verified matches the very lows win,next comes the lows and then the moderates….:)
One of my lows is a known 3rd cousin once removed.
That got me thinking about why the low confidence matches are so important at AncestryDNA.
cM vs Mb
AncestryDNA is a curious animal. Unlike other companies, they phase the data from the get go — meaning that smaller segments are more confident and that they have many fewer IBS segments. This is a much more advanced methodology, but you can’t assume that one advanced methodology means that all of their methods are state of the art. As was revealed in an interview with CeCe Moore on her “Your Genetic Genealogist” blog last year, Ancestry uses a straight Mb comparison. That means, that they don’t convert the segments to centiMorgan (cM) distances. The cM unit of measurement accounts for variation in the likelihood that a segment will be cut in half during meiosis. It’s an empirically derived map that takes into account things like how large the chromosome is, how long the arm of the chromosome is, how far from the centromere it is, and how close to the telomere the segment is. All these physical variations change the odds of a segment being randomly cut up. There are other inhibitors as well, and women cut their DNA up twice as often as males when they are producing sex cells — therefore a sex averaged cM map is used to convert segment size in Mb (mega-basepairs) to size in cM. One cM is roughly equal to 1 Mb in humans but it varies a lot from place to place and chromosome to chromosome.
As Shakespeare would have said, “Aye, there’s the rub!” for small Mb segments can be quite large when converted to cM. If you and your DNA cousin match on one of these frequently cut up segments — often found in the middle region on the long arms of larger chromosomes — then it’s more meaningful and, when converted, would be a much larger cM segment. So a 4-9th cousin — or third once removed cousin — who might only share one single segment with you, may have a very tiny segment identified as ‘very low’ confidence. In reality, when the adjustments are made for the segments location and chromosome via the cM conversion charts, that segment is quite large. Almost certainly, such a segment would be over 15cM and perhaps as large as 50cM!
Why AncestryDNA chose this method is unknown at this time.
Here is an example from build 36 (FTDNA and 23andMe are currently converting to build 37)
Let’s say you have a cousin on chromosome 1 from RS3737728 to RS903903 (These are SNP locations)
|Markers name||Build36 map physical position||Sex-averaged map position||Female map position||Male map position|
As you can see, the results vary depending on the way you look at them:
SO a tiny 1.2 Mb segment is a much larger 4.46 cM (sex averaged) segment. Also note that sex averaged isn’t just an average of the female and male numbers.. shrugs.
AncestryDNA has said they will release the segment information sometime this year. When they do, the information will still not be directly convertible to compare to other companies. You will need to convert the Mbp to cM to be meaningful. But you should be able to do a direct comparison to your AF locations… although we don’t know what build AncestryDNA is on.. one assumes that they started with the latest build that everyone is striving to convert to now. Yeah, yeah, I know. Assuming is how we got here to start with!
Filed under: AncestryDNA