FLICKR, JORIS LOUWESAnalyzing the genetic differences among nearly 2,000 people, James Fowler of the University of California, San Diego, and Yale University’s Nicholas Christakis found that friends shared about 0.1 percent more DNA than the average stranger—a level of genetic similarity expected among fourth cousins.
“Most people don’t even know who their fourth cousins are,” Christakis told BBC News. “Yet we are somehow, among a myriad of possibilities, managing to select as friends the people who resemble our kin.”
“It’s obvious that humans tend to associate with other people who are very similar to themselves,” Matthew Jackson, a professor of economics at Stanford University who studies social networks, told NPR. “This gives us evidence that it’s operating not just at a level of very obvious characteristics but also ones that might be more subtle—things that that we hadn’t really anticipated.”
The data were collected as part of the Framingham Heart Study. “Because the study started in a small community, many people that were named as friends, also happened to be involved in the study,” Fowler told BBC News. The researchers calculated a kinship coefficient for each pair of people based on 500,000 single nucleotide polymorphisms (SNPs) across the genome. Their results, which included the fact that friends tend to have a slightly higher kinship coefficient than strangers, were published this week (July 14) in PNAS.
But some in the field are skeptical about the duo’s conclusions, arguing that some important factors—such as ethnicity and education level, which can lead to population stratification that makes two people more likely to meet and become friends—were not controlled for.
In a second analysis of 907 pairs of friends and nearly 1.5 million SNPs, Fowler and Christakis attempted to control for such stratification or family relationships, but still not everyone was convinced. “I wonder whether [these methods can fully account for] factors known to drive friendships, like church membership, sports or other cultural affinities, that would also lead to a correlation with genotype,” Rory Bowden, a statistician and lecturer at the Wellcome Trust Centre for Human Genetics in Oxford, U.K., told the BBC, “because they reflect differences in places of origin within Europe of the Framingham participants.”
“I think that they’re unusual findings, and that usually draws criticisms from scientists,” Fowler told the BBC. “We’re not really making claims about specific candidate genes here. We’re making claims about structural characteristics across the entire genome.”