#autotree

Genetic Affairs now presents the AutoTree feature. AutoTree compares Ancestry trees to find common ancestors (MRCAs) and reconstructs a tree. AutoTree works for persons with unknown parentage to their birth families (for instance adoptees or donor-conceived persons) as well as people that have linked their DNA profile to a tree. If an Ancestry subscription is available, AutoTree will also be able to use unlinked trees. An Ancestry subscription is therefore highly recommended to get the best results since not all DNA matches are properly linked to trees.

AutoCluster first organizes your DNA matches into shared match clusters that likely represent branches of your family. Everyone in a cluster will likely be on the same ancestral line, although the MRCA between any of the matches and between you and any match may vary. The generational level of the clusters may vary as well. One may be your paternal grandmother’s branch, another may be your paternal grandfather’s father’s branch.

By comparing the linked and unlinked trees from the members of a certain cluster, AutoTree is able to identify ancestors that are common amongst those trees. First, we collect the surnames that are present in the trees and create a network using the similarity between surnames. Next, we perform clustering on this network to identify clusters of similar surnames. A similar clustering is performed based on a network using the first names of members of each surname cluster. Our last clustering uses the birth and death years of members of a cluster to find similar persons. As a consequence, initially, large groups of tree persons (based on the surnames) are divided up into smaller groups using the first name and birth/death year clustering. Last, we also try to identifiy common locations. Locations that are characteristic of the identified clusters can yield information concerning the historical or demographic significance of a cluster. To identify these location clusters, we perform a clustering based on the distances between the entered birth locations of the persons from each of the trees of the DNA matches. Birth locations that are identical or in close proximity (i.e. within a certain radius of meters) are placed into location clusters.

The common ancestors and their descendants are then used to reconstruct a genealogical tree. The tree visualization is based on the WATO visualization and is kindly provided by Jonny Perl (thanks Jonny!). In some cases, quite a few DNA matches can be combined into a single tree (see Figure 1). In practice, several smaller trees are constructed which sometimes can be combined (manually) into larger trees. The linked DNA matches are visualized at the far right using a color gradient to quickly distinguish between different matches with respect to shared cM. Tree persons from unlinked trees are visualized using a brown/yellow color. In some cases, the genealogical path to the tested person will be available in the visualization, you can spot the tested person by its green visualization (see Figure 2). As a visual indicator, we place (Christmas) trees in the AutoCluster clusters and provide information in the popup to indicate if a common ancestor is found between these shared matches (see Figure 3).

The overview of these analyses is displayed in the main AutoCluster HTML file in a table (see Figure 4). For each AutoCluster cluster the number of common ancestors (see Figure 5), common locations (see Figure 6) and common surnames are shown. The fields of the tree, common ancestor and common location are clickable and will show more detailed information.

Last but not least, we calculate the reconstructed trees, common ancestors and common locations as well using all of the DNA matches from our AutoCluster analysis. This ensures that potential shared common ancestors (for instance from two neighboring clusters that are linked by grey cells) are identified between clusters as well.

Please note that, in order to reduce the computational load on the servers of Ancestry, some rate-limiting measures have been implemented (e.g., AutoCluster jobs end up in a queue). The last measure ensures that only a reasonable number of analyses can occur in parallel.