Archiver > AUTOSOMAL-DNA > 2014-04 > 1396586357

From: Don Worth <>
Subject: Re: [AUTOSOMAL-DNA] The significance of segment overlaps
Date: Thu, 3 Apr 2014 21:39:17 -0700
References: <6420670.2501550.1396584058103.JavaMail.root@vznit170158>
In-Reply-To: <6420670.2501550.1396584058103.JavaMail.root@vznit170158>

Thanks, Jim. It seems like we are pretty much in agreement.

One interesting comment you made was that your TGs can contain two segments that do not overlap at all - the first segment(s) in a TG and the last one(s). I see this all the time in my ADSA outputs. The segments "march" across the page from left to right and by the time the later ones start the early ones have ended (if they are short) or, at least, they don't overlap the early ones by much at all. This, of course, affects the ICW outcomes since the early ones don't (necessarily) ICW the later ones. In ADSA the rectangular block of little colored bricks kind of "flakes" away at the upper left corner and the lower right corner so that it's not a perfect rectangle of ICWs.

It feels a little "transitive" to presume that the early ones and the later ones are part of the same ancestor's segment. I guess I could go either way on that. I worry about two very different ancestral-segments lying side by side abutting each other with little match-segments spread out across them like overlapping shingles. If even one segment in the middle doesn't overlap the next (or has fuzzy boundaries) this scenario seems possible.


On Apr 3, 2014, at 9:00 PM, Jim Bartlett <> wrote:

> Don;
> I'll add some of my thoughts and experience [in brackets] in your email
> below
> Jim Bartlett
> On 04/03/14, Don Worth<> wrote:
> I think this question has been asked in various ways in other forums, but
> I'm trying to get a more precise understanding before I start promulgating
> incorrect information to groups that I'm presenting to over the next few
> weeks.
> We all know about John Walden's little table that indicates the likelihood
> that a segment of a given length is IBD or non-IBD (IBS). For example: a 7
> cM segment is 30% likely to be IBD while a 5 cM segment is only 5% likely to
> be IBD. [Tables of this type depend on the algorithm used by the company -
> FTDNA uses a multipart algorithm and has fewer IBS, than say GEDmatch, for
> instance. Someday I may try a table for each company - I am in the process
> of culling out a lot of IBS segments - almost all below 10cM]
> Can anyone opine on these questions on the basis of their experience?
> 1 - Does the fact that two segments overlap and the corresponding matches
> are ICW affect the probability that a segment is IBD? [I definitely think
> so! If several overlapping segments are all on each other's ICW list - I
> think they are Triangulated over 90% of the time] For example, if you have a
> Triangulated Group (TG) with two or more overlapping segments in it that all
> ICW, and you find a, say, an overlapping 7 cM segment that ICWs as well - is
> it more likely than 30% to be IBD? [In my opinion, folks - particularly
> distant cousins - who are ICW on overlapping segments almost have to all
> have IBD segments. Think about the 1,000 or so SNPs in those segments which
> are all the same. When several folks have the same series of segments, they
> got it from the same ancestor]. What about a 3 cM segment that is completely
> overlapped and ICW with a TG? [If it matches with everyone who overlaps it,
> then even a 3cM segment would be in the TG. As you get into the 3cM range,
> however, the chance of an IBS segment grows - shorter segments are just
> easier to fabricate and make into matches - so it's a judgment call. Some
> 3cM segments are IBD, so Triangulation is a good clue as to which one. There
> are NO hard rules (that I know of), however, that MAKE a certain segment be
> IBS or IBD]
> So how small can a segment be and still participate in a TG? [see above -
> use your judgment - which comes from experience - which provides plenty of
> mistakes....]
> 2 - How much must a segment overlap another to be considered "overlapping"?
> (assume the matches involved ICW) Is the area of overlap governed by John
> Walden's predictions? Or do you consider the entire length of each of the
> segments that partially overlap one another?
> [As a general (repeat general) rule, I try to have the overlap meet my 7cM
> baseline for most Matches. Someday, when I'm about 90% mapped and confident,
> I'll go back and see what 5cM segments "fit". Or, on occasion, with a know
> close cousin, or some tantalizing segment, I'll try some 5cM segment to see
> what pops up - I really miss the GEDmatch segment utility. If you start to
> develop a solid TG with 5-10 Matches with segments mostly over 10cM - in
> other words you become confident in the TG - then you can deviate a little.
> If they are all ICW, you can deviate a little more. Use judgment, as, again,
> there are no hard rules that the DNA has to follow. As you look at all
> the overlapping segments in a given area of a chromosome, each one does have
> to obey one rule - they are on one chromosome or the other. I extrapolate
> this like so: each segment should by in TG-A or TG-B or be IBS or be
> UNDETERMINED (there is not enough info to confidently place it in one of the
> other 3 groups - wait: new Matches/segments are coming every week)]
> So how much of an overlap area must there be for a segment to be considered
> overlapping another to form a TG? [7cM]
> I had been working on the assumptions that:
> A. Short segments are more likely to be IBD if they are part of a TG
> (although I don't know that I would go as far as to say a 3 cM segment was
> IBD - especially since the reason you ICW the owner of that segment is
> probably due to a longer segment somewhere else that could be associated
> with a different ancestor). [I generally agree with this]
> B. The overlap must be sufficient to stand on its own as a likely IBD
> segment (although, since it is part of a TG, it might be more likely IBD).
> The size of the portions of segments that don't overlap should not be
> considered when declaring a TG. (Although, presumably they could represent
> differing portions of a longer segment (that you have) from the common
> ancestor.)
> [From all of the segments in a TG, you can make some calculations:
> 1. The size (use Mbp) of the overlap - although I have some TGs that are so
> overlapping, that the "first one" and the "last one" don't overlap at all
> 2. You can calculate the segment you got from the Common Ancestor - Start of
> the "first" shared segment with you to End of "last" shared segment with
> you]
> 3. If you can view the segments the Matches in the TG share with each
> other (at FI:A or GEDmatch), you can calculate the total segment that the
> Common Ancestor had to have, that was passed down the various lines of
> descent to the various Matches]
> [I like to envision two long rows of bricks. The whole row is a chromosome;
> the bricks (would need to be of various sizes) are the segments from
> different ancestors; and the two rows are from your two parents. When you
> form TGs, you are forming the bricks for both rows. So the TGs from one
> chromosome, will not overlap each other, and may have varying gaps between
> them, until you find enough Match/segments to define the whole brick
> (segment from an ancestor)]
> I would appreciate some feedback on this before I make a fool of myself in
> my slide show!
> [Hope this helps - jvb]
> Thanks!
> Don
> ______________________________
> For answers to Frequently Asked Questions about mailing lists, please see:
> -------------------------------
> To unsubscribe from the list, please send an email to with the word 'unsubscribe' without the quotes in the subject and the body of the message

This thread: