#small segments

When it comes to small DNA segments, we’ve heard the “glass half empty” version of the story many times. Here’s the other side of that story.

Submitted for your consideration: A pair of third cousins twice removed and their 6 cM connection…

According to AncestryDNA, Bryan Smith and his cousin, K, share 6 cM of DNA across 1 segment. And according to Ancestry’s ThruLines, Bryan and Cousin K share a pair of third great grandparents, Reuben Willis Smith and his wife Mary Connell.

The cM value is certainly consistent with the identified relationship but did Bryan and Cousin K inherit their shared DNA from the Smith ancestry as shown? Is the 6 cM segment even valid or could it be an artifact of an imperfect DNA matching algorithm?

Let’s start with an easy evaluation: the shared match list.

Among Bryan and K’s list of shared matches at AncestryDNA:

HG, a descendant of Reuben and Mary’s son, Charles Thomas Smith (HG shares 57 cM with Bryan)

RR, a descendant of Reuben and Mary’s daughter, Fannie Janes Smith (RR shares 47 cM with Bryan and 38 cM with K)

IG, another descendant of Reuben and Mary’s son, Charles Thomas Smith (IG shares 47 cM with Bryan and 71 cM with K)

And at least three other descendants of Reuben and Mary are on the shared match list.

So we’re off to a promising start. In addition to the fact that Bryan and K share DNA and a paper trail leading to Ruben and Mary, this group of matches gives us more evidence suggesting that Bryan and K might be related as suspected.

But what about that 6 cM segment shared by Bryan and K? Is it valid? Did it come from the shared Smith ancestors or did it originate elsewhere?

To get the most comprehensive help in answering these questions, we turn to GEDmatch. As indicated in the ThruLine image above, both Bryan and his father are related to DNA Cousin K through their Smith line. And because K is on GEDmatch, we can see that Bryan and his father both share DNA with K on a specific portion of Chromosome 12:

Further investigation reveals that two other descendants of Reuben and Mary, Cousins I and G, share DNA with Bryan and his father on Chromosome 12 in roughly the same location. In fact, all of the matches in question match each other on Chromosome 12:

This is what we call a Triangulation Group. It brings the possible genetic connections into sharper focus.

The common segment shared by all of the members of this Triangulation Group indicates that they all share a common ancestor. And we’ve already identified shared ancestry through the Smith line. Cousins I and G are first cousins once removed and they are descendants of Reuben and Mary’s son Charles Thomas Smith...

A review of the pedigrees of the matches in question reveals no lines of shared ancestry other than the known shared Smith line. This investigation is summarized briefly in the table below, listing 2nd great grandparent surnames and shared ancestors (blue for paternal names and surnames, light red for maternal names and surnames):

Although we cannot say with perfect certainty that there is no additional common ancestry that conceivably could account for the shared segment of DNA on Chromosome 12, the known evidence doesn’t leave room for much doubt.

For completeness, here’s a chart summarizing the amount of DNA shared by the relatives in question:

And cluster analysis for Cousin G yields a cluster with eight descendants of Reuben Willis Smith, including Bryan Smith and Cousin K:

Not everyone will feel the need to go this far to investigate a 6 cM connection. But this post provides examples of ways to investigate the validity of an ordinary small segment and to determine whether the shared DNA legitimately belongs with the presumed paper trail source of the DNA.

Discussion

Skepticism regarding small segments of shared DNA is appropriate. In comparison to larger shared segments, such segments are more likely to be IBS (false).  Additionally, even when small segments can be shown to be reliable, we have to grapple with the fact that small segments can be too old to fall within the reach of reliable historical documentation. 

With the exponential growth of the DNA matching databases, the impetus to explore distant matches waned. Reluctance to do the strenuous work involved in using small segments grew. With access to strong genetic connections leading back to target ancestors, why bother with low cM connections?

The sentiment is understandable!

On the other hand, I believe that excessive skepticism has impeded progress in genetic genealogy.  As databases have grown, our opportunities for research have multiplied and our research techniques have improved. But at the same time, goalposts for small segment success have been moved to poorly-defined and very unreasonable points. 

[From the skeptics: Your success with a small segment doesn’t count if you find a larger segment in a relative! I don’t want to hear about triangulation! Visual phasing is not allowed!]

If we applied such arbitrary restrictions to all areas of genealogy, we’d struggle to get our work done!

Even with our luxuriously large DNA databases, distant genetic connections are the only connections available in some areas of investigation (or to people who hail from less heavily-tested populations). Defeatist refusal to accept low cM matches as evidence in genetic genealogy needlessly limits our potential. 

Don’t get me wrong, I’m fully in favor of scholarly rigor. But let’s not allow skepticism to pave the way for denialism! 

When distant genetic connections are found to be of dubious quality, they should be set aside. But shared segments should not be judged on the basis of size alone. Even the most fervent opponents of small-segment research will admit that small segments are often valid (IBD). And while these opponents frequently cite IBD/IBS percentages, they ironically fail to see that our ability to find these percentages points directly to a practical method for sorting distant matches on an individual basis. 

We are privileged to have access to enormous databases of incredibly valuable genetic information. More than a statistical hiccup that can lead us serendipitously to more reliable information, small DNA segments are  messages we carry with us every day, testifying to our connections with our ancestors. Genetic information, even in small amounts, can be just as valuable as any other form of information. We should be good stewards of that information and we should invest good faith effort in understanding how our distant matches can inform us about our rich ancestral history. 

I’ll close with this analogy for small segments: 

You want some refreshing water but the glass is only half-full. Drink it or toss it out?

As I’ve been digging deeper with visual phasing, I’ve been increasingly impressed with the insights that this method can provide.

In essence, visual phasing narrows down the search for shared ancestry by assigning DNA to a specific grandparent. By combining information gleaned from the DNA test results from three or more siblings, we can eliminate two generations of uncertainty. Furthermore, we can check hypotheses against the results of visual phasing -- either to bolster our prior assumptions or to tear down bad hypotheses and start anew.

What I’d like to do now is to show some examples of what visual phasing can tell us about individual matching DNA segments. In particular, I’d like to show how visual phasing can shed light on the the value of examining the full spectrum matching segments, large and small.

(In this post I will look at the results of visual phasing but I will not discuss how to do visual phasing. Please check the ISOGG wiki for various resources that explain the process.)

I will give two examples.

I’ll start with a simple, relatively unremarkable example. On chromosome 2, I and my siblings have visual phasing results that appear as shown:

By convention, I use 

BLUE for paternal grandfather, 

red for paternal grandmother, 

green for maternal grandfather and 

orange for maternal grandmother. 

In this example we will be focusing on the BLUE because we will be looking at Cousin MJ who is related to my family through my paternal grandfather.

Cousin MJ is my 3rd cousin once removed. AncestryDNA says that we share 37 cM across 3 DNA segments. Ancestry also gives us a Shared Ancestor Hint for our MRCA, Sellers Lee, while putting MJ and me together in a DNA Circle for Sellers Lee. 

I used triangulation and visual phasing to obtain additional insights. On chromosome 2, MJ shares a segment of DNA with me and some of my siblings. This segment very clearly triangulates to Sellers Lee’s parents. For my sibling C, this triangulated segment is a hefty one -- 42 cM! But for me the shared segment is much smaller and visual phasing tells us why.

MJ matches C from 71.1 M to 126.4 M.

If we zoom in on this portion of this chromosome, we can see that visual phasing has revealed some DNA from my paternal grandfather in this range: 

Although my brother C has 42 cM of BLUE here, my sister H has no BLUE.

What about me (J)?

If you look at the visual phasing for my DNA in this range, you can predict where MJ matches me. Look for the BLUE. Do I have any blue from 71.1 M to 126.4 M?

Yes !

I do have BLUE in part of this range, but not all of it. For me the blue ends at about 76.5 M. So we should be able to predict that MJ matches me from  71.1 M to 76.5 M. Is this what we find? 

Yes!

It’s a small segment, but with no adjustments to the default settings in the chromosome browser at GEDmatch, we can see that the segment is there, as expected:

But now we’re at risk of getting into a bit of trouble.

As you can see, the segment I share with MJ is only 6.5 cM. That’s getting dangerously close to the minimum acceptable segment size! Most genetic genealogists rightly advise us to avoid segments smaller than 5 cM. So we’re probably safe with this example, but just barely.

On the other hand, if we had ignored the 6.5 cM segment, we would have ignored a small but valuable piece of evidence that helps to tie this match specifically to one of my four grandparents.

Comparisons with other cousins yielded similar results.

Here’s another example.

CS is my first cousin three times removed. Ancestry tells me that I share 83 cM across 4 DNA segments with CS and we have a Shared Ancestor Hint for our MRCA, Claude Butler:

Claude Butler was my 3rd great grandfather through my maternal grandmother.

On Chromosome 14, I share a 38 cM segment of DNA with CS from 28.7 M to 62.2 M.  Results of visual phasing for me and my siblings on Chromosome 14 are as shown:

Orange represents DNA from our maternal grandmother.

As is my wont, I used my kids’ DNA on Chromosome 14 to assist with some multi-generational visual phasing, the results of which are shown here:

Per my convention, red represents DNA from my kids’ paternal grandmother (my mother). 

As we did with the prior example, we can use visual phasing to predict where CS matches other members of my family.

As noted above, CS matches me from 28.7 M to 62.2 M. Zooming in on the relevant portion of Chromosome 14, you can see where CS should share DNA with my sister H. Look at the ORANGE (maternal grandmother):

Does H have any ORANGE from 28.7 M to 62.2 M?

Yes!

H has ORANGE in most of this range, but not all of this range.

We should expect CS to match H approximately from 33.7 M to 62.2 M. We'll see below if that's actually the case.

Let's zoom in on the relevant portion of Chromosome 14 to see what visual phasing can tell us about where my kids might match CS:

Here we’re looking for RED. If I passed the DNA I share with CS to my children, it will be DNA from my mother, i.e., their paternal grandmother (RED).

Starting from the top, does A have any RED between 28.7 M and 62.2 M?

Yes, RED starts for A at 58.0 M. If I gave to A the DNA I share with CS, it would show up between 58.0 M and 62.2 M.   We'll see below if that's actually the case.

What about Q?  Does Q have any RED between 28.7 M and 62.2 M? 

Yes. It’s all RED in this range! If I gave the DNA I share with CS to Q, it will show up between 28.7 M and 62.2 M. The matching segment should be roughly the same size as my own matching segment.

What about W?  Does W have any RED between 28.7 M and 62.2 M?

Yes, but W got less than Q. For W, red starts with the beginning of the chromosome, but it ends  at about 48.5. We would expect CS to match W from  28.7 M to a 48.5 M.

In summary...

CS should match Q from approximately 28.7 M and 62.2 M.

CS matches me (J) from 28.7 M and 62.2 M.

CS should match H from approximately 33.7 M to 62.2 M.

CS should match W from approximately 28.7 M and 48.5 M

CS should match A from approximately 58.0 M and 62.2 M.

Here’s the verdict, from the chromosome browser (with no changes in default settings):

I think it looks good! All segments are at or near the expected locations.

But now we’re really in trouble. Although visual phasing tells us to expect a small segment for A from approximately 58.0 M and 62.2 M, a segment that falls into this range is only 3.1 cM! This falls well below the generally accepted minimum segment size.

But as with the first example, if we ignore this 3.1 cM segment, we ignore a small but meaningful piece of evidence that helps to tie this match specifically to one of my four grandparents.

Discussion

Let me be clear: I agree with those who urge caution in dealing with small segments. 

Shannon Christmas gives good advice:

Know that some DNA segments are so small they reach “the vanishing point” where their value for genealogy becomes dubious at best. The vanishing point is less than 5 cM... 

For the record, if someone tells you to lower GEDmatch’s default matching parameters to 1 cM and 500 SNPs, don’t.

I agree with this wholeheartedly and I recommend that you read the rest of the blog post. Please follow each of the links.

From the time I started working with genetic genealogy in the latter part of 2012, I ignored DNA segments smaller than 5 cM. Leading genetic genealogists had made it very clear: Smaller segments are far more likely to be false positives -- and even when they’re not false positives, they’re far less likely to be of any use for genealogy. This is not mere opinion. A wealth of evidence and experience back this up.

For most of my work, I prefer to see segments larger than 10 cM. For single segment matches, I prefer to be above 20 cM. But more recently, as I began to do more work with my visual phasing results, I started looking at well-researched individual matches more closely to see if they match my family members where they should match. Then I noticed that some of the anticipated match results just happened to be segments smaller than 5 cM.

Alone, these small segments do NOT form the foundation of any hypothesis. But in some cases they appear to be an important part of the complete picture, particularly in cases in which several family members are used for a visual phasing comparison.*

Final warnings: Let’s continue to be exceedingly cautious with small segments! Do NOT lower cM and SNP thresholds to force the evidence to fit a predetermined conclusion!

EXTRA CREDIT: If you noticed that I (intentionally) skipped over something with the second example, please leave a comment below.

*I’ve also seen some potential for identifying IBS segments with this process.

**Screen shots were modified for anonymity and aesthetics.