#visual phasing

[dnagenealogy]

We need a chromosome browser at AncestryDNA!

Andy makes some incredibly strong points in this video and he makes them very well. It's a huge step in the right direction. I hope he's willing to go further with it. He has one foot on "a chromosome browser isn't needed most of the time…it is not used for the vast majority of your genealogy questions" and the other foot on "chromosome browsers are essential…we need a chromosome browser." I know it's dangerous to take a stand that will rub people at Ancestry the wrong way but if we're going to get a chromosome browser at AncestryDNA, we can't vacillate.

[dnagenealogy]

Can a 6 cM connection be meaningful?

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?

Posted with Bryan Smith’s permission. 17 May 2023

[dnagenealogy]

John Jackson Mullis: Exploring a Distant AncestryDNA Hint

AncestryDNA provides hints for very distant Ancestry. Here’s a hint for my distant cousin, PM:

Seventh cousins once removed! What are the odds?

Probably not terribly good. But a connection between the shared DNA and the shared ancestry is plausible.

The paper trail evidence does appear to be sufficient on both sides. So unless there was an NPE somewhere along the way, my match and I both appear to be biological descendants of John Jackson Mullis. And we do appear to share DNA that might well have come from a distant shared ancestor.  But did our shared DNA come from John Jackson Mullis? 

In accordance with the distant relationship, the amount of shared DNA is modest, only 10.7 cM:

Without digging any deeper, many people would infer that the AncestryDNA hint indicates that shared DNA came from the shared ancestor. Don’t make this mistake! Without additional -- very compelling! -- evidence we must assume that the shared DNA could have come someone else (in this case, perhaps one of my other five hundred 7th great grandparents). We must make this assumption even if the paper trail documentation is excellent and the family trees are very complete.

Some background information on the Mullis family: My maternal grandmother is a Mullis, the daughter of Vernon Earl Mullis. Grandpa Earl was a farmer from Bacon County, GA. He was from a third generation of Mullises born in Georgia. Earl’s great grandfather, Jonathan, was born in North Carolina. Jonathan’s great grandfather, John Jackson Mullis, was born in Lancaster County, Virginia and died in Anson County, NC.

Although PM’s mother is French, his paternal lines, including his Mullis line, are from North Carolina. His great grandmother M. Mullis was a fifth generation North Carolinian.

Digging Deeper

So, what can we do with this AncestryDNA hint? Let’s explore and find out!

Let’s start with AncestryDNA’s Shared Matches list for me and PM:

As you can see, three of the four shared matches have no family tree. But the first shared match, AM, is a descendant of John Jackson Mullis and is my first cousin 3x removed through our MRCA, Jefferson Mullis.

That’s very interesting but it doesn’t prove that the DNA that I share with PM is DNA that came from John Jackson Mullis!  To underscore this point, let’s look at the DNA shared by PM and AM. Fortunately both of them are on GEDmatch, so we’re not stuck at a dead end with the information provided by AncestryDNA:

AM and PM share DNA on Chromosomes 2 and 17 but PM shares DNA with me elsewhere:

The location of the shared DNA is important.  DNA on different chromosomes could have come from different lines.

(Also note that although their Mullis relationship is only a bit closer -- 5th cousins twice removed -- AM and PM share significantly more DNA than me and PM. AM and PM share more DNA than we expect for their shared connection with John Jackson Mullis.)

So even with help from GEDmatch, AncestryDNA’s tools cannot provide a clear or convincing explanation for the DNA that PM and I share.

GEDmatch to the Rescue

Let’s look again with GEDmatch. One of my favorite genetic genealogy tools is the People who match one or both of 2 kits tool. It’s comparable to Ancestry’s Shared Matches feature, but it’s quite a bit more powerful.  Plugging PM and myself into this tool, I got a list with my immediate family at the top followed by AM and his son. (AM’s son was missing from the Shared Matches list at AncestryDNA.)

Here’s what I got when I put the GEDmatch list through the 2D chromosome browser:

In the image above, AL represents my sister. AM and PM were discussed above. JB and JH are siblings who descend from John Jackson Mullis through John’s son Solomon Mullis. JB and JH are my 6th cousins once removed. 

JB (and JH) are on my match list at AncestryDNA:

There is no AncestryDNA Hint for JB or JH because their Mullis line ends with Solomon Mullis’s son, John Culpepper Mullis, in their DNA-linked family tree. Furthermore, despite the shared DNA (42.7 cM for me and 17.5 cM for PM), JB and JH do not show up on my Shared Matches list for PM. Likewise, PM is not on my Shared Matches lists for JB or JH. 

But my Shared Matches list at AncestryDNA for JB did show some interesting connections:

CL: A descendant of  William Charles Mullis, grandson of John Jackson Mullis

DD: My 4th cousin 2x removed through Jacob Mullis, grandson of John Jackson Mullis

EM: A descendant of Margaret Mullis, born in Anson County, NC in 1796

LH: A descendant of Dillie Adaline Mullis born in 1829 in Union County, NC

Although AncestryDNA does not show me a connection between JB and PM, the direct connection is very clear at GEDmatch.

The DNA shared by select members of the Mullis group is noted below:

I’m JL and SS is my sister. JB, PM and AM were discussed above. DS appears to have good DNA connections to Mullis lines but has no DNA-linked family tree at Ancestry. SJ doesn’t know of any Mullis connections but she has no information about her father’s side of the family.

Despite the incomplete paper trail information, each of these matches helps to complete the picture by clarifying the apparent Mullis DNA connections on Chromosome 20:

Visual Phasing

My final process in evaluating this group of Mullis matches was to use visual phasing to confirm or rule out my Mullis grandparent as the source of the shared DNA. Despite not having DNA from my parents, grandparents, aunts, uncles or first cousins, I was able to show that the DNA that I share with my Mullis matches did come from my maternal grandmother, daughter of Vernon Earl Mullis. JB, PM and AM all match DNA that came from the Mullis portions of my family’s copies of Chromosome 20. You can read some notes about this process here.

Conclusions

What is the likelihood of finding a DNA segment shared by a group of distant cousins? Whatever your answer to this question might be, it’s helpful to be aware of such connections when they exist. Depending on your perspective, it’s an either an opportunity to learn more about your ancestry or an occasion to reject improbable match results.

Working with the AncestryDNA hint for John Jackson Mullis, I could only see that I share a small amount of DNA with a cousin (PM) who has John Jackson Mullis in his family tree and one of my Mullis cousins on his match list. But by using the tools and resources at GEDmatch, I was able to expand the picture considerably.

[dnagenealogy]

Visual Phasing with Multiple Generations

Visual phasing can be difficult, so I recommend using all of the resources at your disposal if you decide to give it a try.

I recently became interested in an AncestryDNA hint for an ancestor from my maternal grandmother’s line (Mullis). To evaluate this hint, I decided to employ the technique of visual phasing.

My resources:

My own DNA

My siblings’ DNA

My three kids’ DNA

My wife’s DNA and her parents’ DNA (my kids’ maternal grandparents)

Not available: DNA from my parents, grandparents, aunts, uncles or first cousins.

Fortunately, I was working with Chromosome 20, one of the smallest -- and therefore one of the easiest -- chromosomes.

I started with my kids’ DNA. Using one-to-one comparisons with their maternal grandparents as a sort of “cheat sheet” I  quickly came up with the results below:

Illustration 1

A, Q and H are my kids. The vertical black lines designate crossover points. The horizontal lines at the bottom represent the DNA that they share with their grandparents: Blue, paternal grandfather; Red, paternal grandmother; Green, maternal grandfather, Orange, maternal grandmother.

Using my in-laws’ DNA made this process much easier!

Plugging various matches into the “Multiple Kit Analysis” tool at GEDmatch produced beautiful results.

A triangulated segment shared with my paternal third cousin twice removed (3C2R) lined up exactly as expected, matching Q only:

As revealed by visual phasing, Q is the only one of my children who inherited DNA from my father on this portion of Chromosome 20 (see illustration 1).

That was just a little test comparison. Getting back to the original goal, I compared my possible Mullis DNA match with A, Q and H to see whether the results would line up in accordance with the suggested line of inheritance.  If the shared DNA came from from my Mullis line, my kids’ match results would show that the DNA came from their paternal grandmother (my mother).

Here’s how the Mullis match (PM) matches me (JL):

On this portion of Chromosome 20, Q and H inherited DNA from my mother but A did not (see Illustration 1). Therefore, PM should match Q and H only.  Here’s what I got when I compared PM with A, Q and H:

Nice!

Another Mullis match in the same location gave similar results...

Like JM, JB should match Q and H if the DNA that I share with him came from my Mullis line. Because of the length of the segment, he should also match A, but that segment should be much smaller as it would start at A’s crossover point at about 54,000,000 (see Illustration 1). Here’s what I got when I compared PM with A, Q and H:

Beautiful!

Emboldened by these results, I phased my own grandparents’ DNA on Chromosome 20 using my DNA results in comparison with the DNA results of two of my siblings:

Illustration 2

As in Illustration 1, the horizontal lines at the bottom of Illustration 2 represent DNA shared with grandparents: Blue, paternal grandfather; Red, paternal grandmother; Green, maternal grandfather, Orange, maternal grandmother.

Anticipating the results we should expect with JB, note the fact that my maternal Chromosome 20 appears to match my maternal grandmother (Mullis) from beginning to end. But you will also note that there is a crossover point at about 41,000,000 for S and a crossover point for L at about 48,000,000.  If the DNA that JB shares with me came from my Mullis line, JB should share DNA with S, L and me on chromosome 20. But the segment shared with S should be slightly shorter than mine and the segment shared with L should be significantly shorter than mine, in accordance with the locations of S and L’s crossover points.

Here’s what we find:

Perfect!

Careful analysis of the results of my multi-generational visual phasing project appears to show that I’ve successfully sorted my DNA matches on Chromosome 20 by grandparent and that my Mullis matches share DNA with my family that came from our Mullis line.

This process is so powerful it feels like cheating!