Last month at RootsTech 2023 we announced the release of cM Explainer, a powerful feature that estimates familial relationships between DNA Matches. For every DNA Match, cM Explainer provides predictions of possible relationships between the two people, and their respective probabilities. cM Explainer uses a novel approach that calculates the relationship predictions using the ages of individuals, in addition to the amount of shared DNA. As with all MyHeritage features, we’re constantly refining cM Explainer to improve its accuracy. Today, we’re happy to share several enhancements we’ve made to cM Explainer, to improve its prediction accuracy for distant DNA Matches.
Improved predictions for distant DNA Matches
DNA Matches on MyHeritage appear, by default, in descending order, from the greatest amount of shared DNA to the least. When a small amount of shared DNA is present, it becomes more difficult to distinguish between distant relationships. As the number of possible relationships increases, it can become even more difficult to predict the exact relationship between two individuals.
We recently fine-tuned our prediction model to better handle low centimorgan (cM) values. As a result, some DNA Matches will now receive predictions indicating a more distant relationship than before. For example, matches who were previously predicted as probable 3rd cousins may now appear as probable 4th cousins, and 4th cousins may now appear as probable 5th cousins. This enhancement to the cM Explainer model applies to relationship predictions for DNA Matches on MyHeritage, and also when analyzing matches using the standalone cM Explainer tool.
Another change we’ve made is that for all matches with 100 cM or less of shared DNA, we now refer to the relationship prediction as an “estimated relationship” rather than a “probable relationship.” We made this change because predictions are inevitably less accurate for low centimorgan values.
More accurate predictions for distant matches of Ashkenazi Jewish origin
Among endogamous populations, where there is a high incidence of marrying within the same community, two individuals may have shared DNA from multiple common ancestors. This results in a larger amount of shared DNA than is expected for their relationship. Originally, cM Explainer did not take into account the ethnic origins of the DNA Matches when calculating relationship predictions, which meant that the probable relationships between some matches appeared closer than they really are.
As part of this update, we’ve fine-tuned the relationship predictions for DNA Matches on MyHeritage with Ashkenazi Jewish ethnicity, due to the history of endogamy within this community. This adjustment is applied to distant DNA Matches with fewer than 200 cM of shared DNA. cM Explainer now uses a formula developed by our Science team that considers the degree of Ashkenazi Jewish ethnicity within a DNA Match, and adjusts the shared DNA used by the algorithm to offer a more accurate relationship prediction. The more Ashkenazi ethnicity the two people share, the greater the adjustment to the algorithm. MyHeritage is the first and only company to provide relationship predictions that account for endogamy. This adjustment is not available in the standalone cM tool because the tool does not take ethnicity data into account.
In the example below, one match is 44 years old and has 90.7% Ashkenazi ethnicity, and the other is 54 years old and 96.4% Ashkenazi, and they have 70 cM of shared DNA. The original relationship prediction for this match indicated a probable 3rd cousin relationship. Following the general improvement to the prediction model, together with the added adjustments for two matches with significant Ashkenazi Jewish ethnicity, cM Explainer now predicts a 5th cousin relationship and calls it “estimated” rather than “probable.”
We may fine-tune the algorithm in the future to provide more accurate predictions for additional endogamous populations.
We are continuously enhancing cM Explainer to improve the accuracy of relationship predictions. We hope these improvements enable you to more efficiently manage your DNA Matches and better understand family relationships.
Source: My Heritage