One of the concerns about using genetic data along with medical records information to personalize medicine is how to keep that personal information safe, but still easily accessible for analysis. Cryptographers at a workshop hosted by the University of California, San Diego, tested a homomorphic encryption method that seems promising, reports Nature News' Erika Check Hayden.
This method involves mathematically encrypting data on a local computer and then uploading the encoded form to the cloud where it can be analyzed, Check Hayden notes. Encoded results are then sent back to a local computer, which unscrambles the data. Any data intercepted along the way would be encrypted.
She notes that this idea dates back to 1978, but remained largely theoretical until 2009 when IBM Thomas J. Watson Research Center's Craig Gentry showed that computational analyses could be carried out on homomorphically encrypted data.
At the UCSD workshop, cryptographers showed that such an approach could analyze data from 400 people within about 10 minutes and pinpoint a variant associated with disease from among few hundred loci. Analysis of larger datasets and more base pairs wasn't always possible, Check Hayden says, and it could take a lot of computer memory, time, or money.
While the workshop organizers find the approach promising, others say it might not provide enough protection for the data or allow researchers and clinicians to perform all the analyses they want. US National Center for Biotechnology Information's Steven Sherry, for instance, prefers restricting data access to a select few people who have agreed to follow certain regulations on how the data may be used.