How to solve an unknown parentage case with DNA

Before you start

An unknown parentage case is any search for a biological parent the person doesn't know: an adoptee looking for birth parents, a donor-conceived person, or someone whose DNA results revealed an unexpected parent (an NPE). The same steps find an unknown grandparent.

The method is the one investigative genetic genealogy uses on cold cases: find the people who share your DNA, build their family trees until they meet, then trace that family forward to the person who fits. It works best with at least one match above about 90 cM on the side you are searching.

The nine steps

1. Test, and transfer where you can

   Test with the company that has the most testers (Ancestry, then 23andMe and MyHeritage), and upload the raw data to GEDmatch, MyHeritage and FamilyTreeDNA, which accept transfers. Every extra database is another chance of a close match.

2. List your closest matches

   Write down every match above about 30 cM with their shared cM, the longest segment if you have it, and their shared matches. A half sibling, aunt or first cousin can solve a case on their own; most cases start with second to fourth cousins.

   In CanopyLink: Match list import turns a CSV from GEDmatch, MyHeritage, FTDNA or 23andMe into one tree per match, with their cM.

3. Sort matches into the maternal and paternal sides

   Use shared matches to group them: matches who match each other usually descend from the same ancestral couple. The Leeds method or an automated tool such as Genetic Affairs' AutoCluster gives you four groups, one for each grandparent, if the data is good. Known relatives on one side, such as a mother who has tested, take whole groups off the table.

4. Build each match's tree back to their great-great-grandparents

   For each match in a group, build their tree from public trees, obituaries, census records and vital records until you reach the generation your cM suggests you share. Record a source for every relationship you add, because a wrong link here sends the whole case the wrong way.

   In CanopyLink: Each match gets a tree of their own in the case project, with sources, a timeline per person, and duplicate warnings as you type.

5. Find where the trees meet

   Look for the couple that appears in the trees of several matches in the same group: the most recent common ancestors (MRCA). Shared surnames and shared places across trees are the first clues.

   In CanopyLink: Link the trees at the shared ancestors, color each match's Match Line, and the shared ancestors are marked where the lines meet. Match Analysis lists shared surnames, places and likely duplicates across trees.

6. Build the common ancestors' descendants forward

   Trace every child, grandchild and great-grandchild of the common ancestors down to the present day. This "reverse" or descendancy genealogy is where the candidates appear: the people of the right age, sex and place to be the unknown parent.

   In CanopyLink: Descendants From Here shows the common ancestors at the top with all their downlines, and the Project Map shows where each line lived.

7. Test each candidate against the DNA

   For each candidate, check that the expected relationship to every match fits that match's shared cM using The Shared cM Project ranges. DNA Painter's What Are the Odds? scores competing hypotheses side by side. Rule out candidates whose ages, locations or cM don't fit, and write down why.

   In CanopyLink: The built-in shared cM calculator shows every plausible relationship for a cM value.

8. Confirm with targeted testing

   A genealogical conclusion is a lead, not a proof. Confirm it by asking a close relative of the candidate to test, such as a half sibling or the candidate's own child, and checking that the result matches what the tree predicts.

9. Write it up

   Record the tree that supports the answer, the sources behind each link, the candidates ruled out and the confirming test, so the conclusion can be checked by someone else.

   In CanopyLink: Candidate reports draw a PDF around the candidate, number the matches instead of naming them, and mark the common ancestors.

Common pitfalls

• Copying public treesOther people's trees are clues, not sources. Check each link against a record.
• Reading cM as one relationshipMost cM values fit several relationships. Keep every one in play until the tree rules it out.
• Duplicate peopleThe same ancestor built twice in two trees hides where the trees meet. Merge them once you're sure.
• Endogamy and pedigree collapseCommunities that married within themselves share more DNA than the tree suggests, so expect higher cM.

Comparing tools for this work? Read IGG software: the tools, explained.