For Southeast Asians and tropical Pacific islanders, the algorithm matched 87 percent to the island where their ancestors originated. And for a group of people from the Italian island of Sardinia, the algorithm matched 25 percent to their ancestral village, and the remainder to within 31 miles (50 km) of their villages.
For comparison, the researchers tested the best alternative algorithm, developed by researchers at the University of California, Los Angeles, on their data and found it identified the correct country of ancestral origin for only 2 percent of people.
The genetic test was less accurate for highly mixed populations, such as Bermudans and Puerto Ricans, who have experienced major demographic shifts over the past few hundred years, Elhaik said.
For people of mixed ancestry, whose parents originated in different places, the algorithm predicted the geographic midpoint of both places. In the next version of the algorithm, called GPS2, the researchers aim to predict the country of origin of each parent.