Wednesday, July 31, 2019

Interpreting GEDmatch AYPR results

If a child inherits the same allele from both parents at the exact same location, on both the paternal chromosome and the maternal chromosome, they are said to be homozygous at that location. When parents are biologically related they can share strings of alleles inherited from common ancestors, and they can both pass on the shared string or part thereof to a child. These homozygous segments are known as runs of homozygosity (ROH) and GEDmatch has a handy tool to calculate the extent of such segments in our DNA. This can be helpful when you don’t know who one or both of your biological parents are, as a) having related parents makes sorting DNA matches into genetic networks and predicting relationships to DNA matches harder so it is best to be forewarned, and b) it gives you useful information if you already know who one parent is.

The meaning of total shared segments reported by the GEDmatch Are Your Parents Related? (AYPR) tool is not obvious. Some genetic genealogists have pointed out that multiplying the total by a factor of four approximates the amount of DNA shared by the child’s parents, from which we can infer some possible relationships between unknown parents (see Homozygosity on the ISOGG Wiki for blog links). I wanted to try explain why, to myself and others.

The first column of figures in the table below will be familiar to genetic genealogists used to working with autosomal DNA statistics associated with the theoretical coefficient of relationship for different relationships (A). The last column of figures containing theoretical total homozygous autosomal segments of the child (H) is based on the theoretical coefficient of inbreeding for different relationships (half the relationship coefficient). I have tried to keep this explanation as simple as possible, assuming an otherwise outbred ancestry. I have listed my scientific sources at the end for those who want to know more about the underlying theory based on Mendelian genetics.

In the table, L is shorthand for the total length of the 22 autosomes in centiMorgans (cM). Because we have two of each chromosome, shared DNA percentages apply to double that length i.e. 2L. The theoretical proportion of a child’s autosomal genome length (L) that will be homozygous is the child’s inbreeding coefficient. We can convert these percentages to cM using the appropriate factor, 2L or L. The GEDmatch AYPR result corresponds to the figures in the last column (H).


Degree of parental relationship with examples
Theoretical total shared autosomal DNA of parents
(A)
Theoretical total homozygous autosomal segments of child
(H)
First (parent/child, full siblings)
50% x 2L
100% x L
25% x L
Second (half-siblings, uncle/niece, grandparent/grandchild, double first cousins)
25% x 2L
50% x L
12.5% x L
Third (first cousins)
12.5% x 2L
25% x L
6.25% x L
Fourth (first cousins once removed)
6.25% x 2L
12.5% x L
3.125% x L
Fifth (second cousins)
3.125% x 2L
6.25% x L
1.5625% x L


Consider the example of half-siblings M and F below, where their shared parent P has a section of DNA at a specific chromosomal location that I will call pq, the p half inherited from one of their parents and the q half from the other. Two out of four (50%) possible ways M and F can inherit DNA from P at this location result in them being identical by descent (IBD) at that location. I will call the IBD half i and the half each child inherits from their other parent j and k. One out of four (25%) possible ways a child C of M and F can inherit DNA from both parents at this location results in them being homozygous at that location.

Parent P (pq)
IBD probability
Half-sibling parents
M (ij) and F (ik)
ROH probability
Child M
Child F

Child C

p
p
25%
i
i
25%
q
q
25%
i
k
p
q
j
i
q
p
j
k


Total 50%


Total 25%

Half-siblings M and F share 50% x L (25% x 2L)

ROH segments of their child = 50% x 25% x L = 12.5% x L

In the parent/child or full siblings scenarios, the parents share 100% x L (50% x 2L)

ROH segments of their child = 100% x 25% x L = 25% x L

The autosomal DNA shared by the child’s parents can be approximated as 4 x the child’s total homozygous segments reported by GEDmatch.

Note that these figures are theoretical averages for random processes, not the ranges we observe in practice. Also note that a child's parents may still be related within a genealogical time frame if the GEDmatch AYPR result is zero, as they may not both pass on the same segments above the threshold used by GEDmatch to eliminate false positives.

If you need support dealing with an unexpected positive AYPR result please read the High ROH Infosheet prepared by genetic counselor Brianne Kirkpatrick.

Sources:

Franklin, I.R. “The distribution of the proportion of the genome which is homozygous by descent in inbred individuals.” Theoretical Population Biology 11, no. 1 (1977): 60–80.

Sund, K.L. and C.W. Rehder. “Detection and reporting of homozygosity associated with consanguinity in the clinical laboratory.” Human Heredity 77, no. 1–4 (2014): 217–24.

Sund, K.L et al. “Regions of homozygosity identified by SNP microarray analysis aid in the diagnosis of autosomal recessive disease and incidentally detect parental blood relationships.” Genetics in Medicine 15, no. 1 (2013): 70–78.

Thompson, E.A. "Descent-Based Gene Mapping in Pedigrees and Populations." In Handbook of Statistical Genomics, 4th edition. David J. Balding, Ida Moltke and John Marioni, editors. Hoboken, New Jersey: John Wiley & Sons, 2019. Chapter 20.

Weir, B.S. “Inbreeding.” In Encyclopedia of Biostatistics, 2nd edition. Peter Armitage and Theodore Colton, editors. Hoboken, New Jersey: John Wiley & Sons, 2005.

Wright, Sewall. "Coefficients of Inbreeding and Relationship." The American Naturalist 56, no. 645 (1922): 330–38.

Monday, January 15, 2018

Desperately seeking Susan: Part 6

(Continued from Part 5)

Within a year of starting her DNA search for her biological family, Bonnie* matched two maternal aunts Bridget* (Part 3) and Grace* (Part 4), and her birth father (Part 5), and identified her birth mother Gail* (Part 5). She has now reunited with both birth parents in person.

In JANUARY 2018 Gail received her AncestryDNA results, confirming their parent/child relationship.

Gail in Bonnie's match list – DNA confirmation

Gail also matched Grace as expected, and asked for my help understanding what their match means.

Gail in Grace's match list – note her Italian ethnicity

In Part 4 Bonnie's search inadvertently revealed that her birth mother was a subject of misattributed paternity, also referred to as a non-paternity event (NPE). My working hypothesis was that Bonnie was the daughter of Grace’s paternal half-sister, who was unknown to Grace. (Incidentally, there was also a recent NPE on Bonnie's paternal side. Even if we could have identified the ancestors of her highest paternal match at 23andMe, a second cousin, it would not have led us to her birth father.)

Possible relationships for matches in AncestryDNA's CLOSE FAMILY – 1st COUSIN category include half-sibling, full aunt/uncle, full niece/nephew, grandparent/grandchild, and double first cousins sharing all four grandparents.

AncestryDNA does not list the half-sibling possibility under actual DNA results

People with any of these relationships share on average 25% of their DNA, and it is not obvious from shared autosomal DNA amounts alone which relationship is the correct one, so we also need to make use of any other information available, such as:

– We can exclude relationships that are not possible based on circumstantial evidence (ages and other tree data).
– We can analyze shared X-DNA at GEDmatch to check whether possible paternal half-sisters satisfy the requirement that they share their father's full X chromosome.
– We can analyze shared autosomal DNA at GEDmatch to check whether matches share fully identical segments, which double first cousins sharing paternal and maternal grandparents are expected to share.

Grace and Gail were born in the same area only four years apart and neither has any full siblings to my knowledge, so we can exclude the grandparent/grandchild and full aunt/niece possibilities. They have unrelated known mothers and don't share fully identical segments (established by a one to one autosomal comparison with graphics at GEDmatch), so we can also exclude the double first cousin possibility.

Gail and Grace share their father's X chromosome

A one to one X-DNA comparison at GEDmatch confirmed they share a full X chromosome, consistent with my working hypothesis and the only remaining possible relationship for their DNA match, paternal half-sisters. Furthermore, their shared matches belong to the tree of Grace's known father Carlo* Jr, indicating that they are both his biological daughters.

Gail's biological father was not the person she thought it was, and none of her parents are alive to ask about it. It is a hard pill to swallow. I can explain what the DNA is saying, but any new information revealed must ultimately be processed emotionally by those affected in their own time.

*This is a true story. I have permission to blog about the people who asked me for help but have not used their real names for the sake of their privacy. I have tried to limit my writing here to information pertinent to their DNA searches, but have shared other details I found with them in tree format.

{Related resources can be accessed via ISOGG's Wiki page on DNA testing for adoptees.}

Friday, January 5, 2018

Desperately seeking Susan: Part 5

(Continued from Part 4)

Holy smokes!

Bonnie's* birth father had already tested with AncestryDNA, and he definitely wasn't Italian (remember my working hypothesis that Bonnie's Italian DNA came from her maternal grandfather Carlo* Jr, Grace's* father and Bridget's* paternal half-brother). His family was from North Carolina and Tennessee. To my horror, I woke up to discover that in her excitement Bonnie had made contact with his daughter (the match admin) immediately, before I had a chance to research him. Fortunately my panic was unfounded as he and his family took the news remarkably well and have been kind and receptive.

In SEPTEMBER 2017 Bonnie met with her birth father and paternal siblings in person.

Unfortunately her father couldn't remember much from the time of Bonnie's birth, but he did think of one person who might be Bonnie's birth mother. Her first name was Sue. We tried to find Sue by name alone, but that is easier said than done when more than forty years have passed and her surname could have changed more than once since then. I realized fairly quickly it wasn't going to work, and she may not even be Bonnie's birth mother anyway. After the excitement had worn off, it was back to the DNA drawing board for me.

Hmmm

Bonnie's closest match in her maternal grandmother's line (i.e. not shared with her father or maternal Italian matches) only shared 70 cM at AncestryDNA. Her test was managed by someone else, and she didn't have any ancestry information visible on her profile. To make matters worse, based on their shared matches she appeared to share French Canadian ancestry with Bonnie, which means endogamy, which means she could be even more distantly related than the shared cM suggested due to multiple connections.

The French and Polish matches had ties to Massachusetts, the French part came from Canada before that, and the Polish part came from recent immigrants. They came in the 19th and early 20th centuries to work in the area's thriving jute mills. Ironically, along with Irish immigrants, they gave New England a strong Roman Catholic presence. (According to Bonnie's non-identifying information, her birth mother was Roman Catholic.)

I was not optimistic about finding Bonnie's birth mother. I kept plugging away at Bonnie's Massachusetts matches on and off, thinking it was impossible but going through the motions anyway. I gave up a few times, checking Bonnie's AncestryDNA match list every day for a new close match that would put me out of my misery.

I worked all the clues the matches offered, grouped the matches based on shared matches, built trees for them, and found shared ancestors with shared matches for a few (this process is called tree triangulation or pedigree triangulation). I added all the bits and pieces I could find to separate "islands" for connected matches in my single research tree space – puzzle pieces of Bonnie's own family tree.** I built the lines of shared ancestors forward and looked for any connection to the other puzzle pieces or to Albany, NY. Working backwards most shared ancestors with Bonnie would have lived in the old countries, which often makes the task of finding connections impossible, and working forwards many tree lines fizzled out as people left Massachusetts for greener pastures and their trail went cold.

One day, while working on a French Canadian branch I had previously thought too close to the 70 cM match for their shared DNA, I happened upon an obituary dated 1959 some kind soul had uploaded to their public tree which mentioned a surviving son who then lived in Albany, NY. Other sources revealed he was married to a woman of Polish descent. Paydirt.

I used obituaries, city directories, newspaper articles, and other sources of information on living people to flesh out his children and grandchildren. One of his daughters, who was a toddler in Massachusetts in the 1940 census, had a daughter Gail* who fit the description of Bonnie's birth mother recorded by the adoption caseworker. Her first name wasn't in the same vicinity of the alphabet as Susan, but Bonnie's birth father confirmed he had in fact known someone with her name. To my amazement, based on shared ancestors in multiple lines and circumstantial evidence, that 70 cM match I once scoffed at appeared to be Bonnie's full second cousin once removed (which is statistically possible). Her grandmother was the sister of the man who had moved with his family to Albany, NY by the 1950s, now believed to be Bonnie's great-grandfather.

In NOVEMBER 2017 Bonnie wrote a letter to her putative birth mother Gail and received a prompt reply confirming that she was in fact the right person, and she was happy to be found. Their relationship has been confirmed by further DNA evidence, and Bonnie will soon be reunited with her birth mother in person.

Birth parents and siblings are not always happy to be found. Bridget's and Bonnie's searches had many twists and turns, but the responses and assistance from members of their biological family including Grace, who tested and shared her DNA data, have more than made up for it.

*This is a true story. I have permission to blog about the people who asked me for help but have not used their real names for the sake of their privacy. I have tried to limit my writing here to information pertinent to their DNA searches, but have shared other details I found with them in tree format.

**I build private and unsearchable speculative research trees on Ancestry's website. More often than not I can find the information I need to build trees for matches without anyone contacting them, which is a frustrating and risky process. I usually temporarily add a match as a new child of any existing person in the tree, then edit relationships in their profile removing the temporary relationship without removing any people from the tree overall. Later on I can link them up to the puzzle wherever they actually fit. I use two different symbols in the name suffix field to denote shared ancestors and paths to them.

{Related resources can be accessed via ISOGG's Wiki page on DNA testing for adoptees.}

(Continued in Part 6)