Experts say genetic privacy fears unfounded after sketchy paper claims DNA IDs physical traits
A new research paper suggests people’s physical traits, such as their face, could be identified through their DNA, but scholarly critics contend that the claims made by the paper are exaggerated, and risk stirring up unfounded fears about identifying people through public genome data.
Published by genome sequencer Craig Venter at his San Diego-based company Human Longevity, Inc., the paper “Proceedings of the National Academy of Sciences” claimed to identify, with 74 percent accuracy, people from a group of 10 people based on small differences in their DNA sequences.
HLI said it developed artificial intelligence to identify differences in DNA associated with facial features, such as cheekbone height, as well as height, weight, and skin color.
But geneticists who reviewed the findings told Nature they feel the concerns expressed are blown out of proportion. In fact, experts conducted similar tests and found that basic information such as age, sex and race would produce the same level of accuracy.
In short, people could be identified with a similar degree certainty without access to any information about their genome at all.
Even biologists who worked on the paper have soguht to distance themselves from it, saying that HLI misrepresented the data from the study. One co-author, Jason Piper, said he believes sharing information about the genome, not withholding it, will lead to more scientific breakthroughs.
The key challenge, Piper said, is finding a way to make genome data public while making it impossible to identify the people associated with the DNA.
HLI, however, defends its findings, and issued a statement saying it intended to encourage discussion on a “sensitive topic” that triggers “emotional debate.” But it hopes that the publication and ensuing discussion will help to emphasize genome privacy protection strategies when such data is exchanged.
“Our work is a proof of concept done with a small set of individuals that presents a strategy for identification of genomic data using predictive modeling,” the response reads. “While facial images provide rich data from which to build phenotypic models, it is important to notice that it is not only faces that may be derived from the genome codes for a vast number of heritable phenotypes.”