A New Coverage Estimator at DNA Painter

Coverage is the biggest concept in genetic genealogy that you’ve probably never heard of.  It refers to how much of an ancestor’s autosomal DNA you can “recover” by testing their descendants.  Each descendant only has a fraction of the ancestor’s DNA:  50% of a parent’s, about 25% of a grandparent’s, and so on.  It adds up, though, when you test multiple descendants of that person, like constructing a mosaic image.  It adds up faster when you test the right relatives.

Coverage matters for genealogy because we won’t match all of the DNA relatives that our parents, grandparents, and so on would have.  A relative you don’t match is a relative you might not know to collaborate with or to use in DNA-based proof arguments.  However, some of our other relatives will have inherited different bits of that ancestor’s DNA and might have key matches we don’t.

Coverage is also important for ancestral reconstruction methods, such as those offered by Borland Genetics.  The more unique bits of ancestral DNA represented in the descendants, the better the reconstruction.

Paul Woodbury of Legacy Tree Genealogists first introduced the idea of coverage to the public in a blog post in April, 2018. He’d been teaching about it in genealogy courses for some time before that.  And while it’s a brilliant approach to devising a DNA-based research strategy, it hasn’t gotten the attention it deserves because it’s quite math-y.  Here is Paul’s equation for someone whose five children have tested:

Yikes!

The calculations become even more complex when you add more testers, when they are in different generations, or when they include mixtures of siblings and cousins.

More recently, Nicole Dyer at Family Locket shared a semi-automated coverage calculator based on Paul’s formulas.  It was created in Google Sheets by Laura Clark Murray.  It’s a huge time-saver because it does the math for you, but it is still based on the complex formulas.

 

A More Intuitive Approach

It turns out there’s a simpler, more intuitive way to evaluate coverage.  Here’s a straightforward example:  A child inherits 50% of their mother’s DNA, so testing the child recovers half of her genome.  A second child also inherits half of her DNA, but in that case, 50% plus 50% does not equal 100%.  That’s because siblings share DNA.  Some of Mom’s DNA in the second child has already been recovered by testing the first child.  In fact, testing the second child recovers, on average, only half of what’s missing, or 0.5 x 0.5 = 0.25 = 25%.

The same logic applies to each additional sibling:  they recover about half of what’s still missing from the previous tests.  This is an average rather than a precise figure, of course, because some siblings share more DNA than others.

As you can see, there are diminishing returns.  At some point, testing more siblings might not be worth the expense.  And since most of us aren’t made of money, coverage is an important consideration as we plan our research strategies.

The issue becomes more important the further back in your tree you go.  Consider the siblings above in relation to their grandfather.  Testing five siblings recovered ≈96.9% of their mom’s DNA but just ≈48.5% of Grandpa’s.  That’s because mom inherited half of his genome.  In fact, even if Mom herself tested, she would only recover 50% of his DNA.

If your objective were to recover Grandpa’s DNA, and none of his children are available, it turns out you’re better off testing five first cousins than five siblings.  Each grandchild only has ≈25% of his DNA, but it adds up.  The first grandchild contributes ≈25%, the second ≈25% of the missing ≈75% (or ≈18.8%), and so on.  With just three first cousins, you could recover more of Grandpa’s DNA (≈57.9%) than with the five siblings (≈48.5%).  And for the same amount of money, you could recover ≈76.3% of his DNA with the five-cousin strategy.

Let’s consider one more scenario:  five grandchildren where two of them are siblings.  We can calculate coverage for the first three grandchildren as before, but for the last two, we need to account for the fact that they are siblings.  Two children cover ≈75% of Mom, who has 50% of her father’s DNA, so those last two grandkids only cover 37.5% (half of 75%) of the missing 42.1%, or 15.8% between the two of them.

Easy peasy, right?

Now you try one:

Ha!  Just kidding.  It’s not easy at all.  In fact, the calculations are a total pain in the patootie.  That’s why we need a calculator that’s both infinitely scalable and intuitive to use for the everyday genealogist.  

 

Introducing the “Coverage Estimator” at DNA Painter

The Coverage Estimator allows you to quickly build or import a tree, mark who in the tree has tested, and see the calculated coverage for the ancestor you are researching.  (We call that ancestor your “research subject.”)

Here’s our Thomas example from above.  He’s estimated to have 52.5% coverage from his tested descendants.

If you recognize the backbone of the What Are the Odds tool here, the Coverage Estimator should be fairly intuitive to use.

For detailed, step-by-step instructions on how to use the tool, see Jonny Perl’s blog post here.  Briefly, you build or import a descendant tree, starting with the ancestor whose DNA you are trying to recover, like Thomas.  The tree should include each descendant who has taken a DNA test.  Those get marked as such using a pull-down menu.

The Estimator does its magic on the fly as you set up the tree.  It’s neat to see the coverage total increase as you add testers.

Best of all, the Coverage Estimator recommends the best person (or people) to test next.  This feature works best if your tree includes death years; the tool will not to recommend someone it knows to be deceased.  In Thomas’ case, it suggests testing either Joanne or Marianne to bring coverage up to ≈58%.

You can even note whether the suggested people are willing to test.  If someone declines, or if they agree but their results are not available yet, the tool will make more suggestions.  That option is found under Add/Edit details for each living person in the tree.

When to Use the Coverage Estimator

Coverage is a concern any time you can’t test your research subject and are working with the DNA of their descendants.  Your research subject is the ancestor whose parentage you are investigating.

For example, several years ago I confirmed the biological parents of my great grandfather, Claude Duval LaCoste, using the DNA results of his three grandchildren.  Claude was my research subject.  The tool tells me that the three tests provided about 53% coverage of his DNA.  Fortunately, that was enough!  The tool can’t suggest additional testers in this case because there’s no one else to include.

Paul Woodbury, the originator of the coverage concept, describes five ways to use the Coverage Estimator.  Check it out!

 

Key Points to Remember

There are a few things to keep in mind when using the Coverage Estimator.

• Consider coverage for every historical project.  That is, any research project that attempts to identify or confirm the parentage of an ancestor who can’t be tested themself (e.g., grandparent, great grandparent, etc.) should include an analysis of coverage.  It’s not necessary to identify your own biological parent(s); your own DNA test represents 100% coverage for yourself.
• The coverage percentages are just estimates.  The true amount of an ancestor’s DNA represented by their tested descendants will vary because inheritance is random.  In other words, the existing tests may cover more or less of the ancestor’s DNA than the tool reports, and it can’t guarantee how much a suggested tester will add.
• Coverage is specific to each database.  If you’ve tested at AncestryDNA and your sister tested at 23andMe, you still only have 50% coverage of your parent in each database.  To accrue the benefits of coverage, the descendants should be in the same DNA database and you should have access to those results.
• Use the Coverage Estimator to guide your next steps.  An analysis of coverage can help you decide whom to test next, whom to share DNA access with, and whom to recruit to a secondary database like MyHeritage or for ancestral reconstruction.  That is, if you are at AncestryDNA and your sister is at 23andMe, if you both transfer to MyHeritage, you will have 75% coverage of your parent there (but not in the two original databases).