MIT, Stanford Project Protects Security of Genomic Data for Open Research

In a paper appearing in the journal Nature Biotechnology, researchers from MIT and Stanford University have described a new system they've developed for protecting the privacy of people who contribute their genomic data to large-scale biomedical studies. These studies are intended to uncover links among genetic variations in identifying the causes for diseases.

As the researchers explained, most sequenced genomes are currently kept in "strict access-controlled repositories." Giving free access to the data through "association studies" could speed up the research. Yet, people concerned about the data privacy of their genetic make-ups may refrain from contributing their genomes to scientific studies. For example, one expert claimed to be able to analyze raw genomic data to determine the shape of faces; and researchers have shown how to triangulate genomic information with other data to elicit the identity of somebody.

The protocol, developed by MIT's Hyunghoon Cho and Bonnie Berger and Stanford's David Wu, is intended to help make currently restricted data available to the scientific community, potentially enabling secure genome crowdsourcing while still making sure individuals can contribute their genomes to a study without compromising their privacy.

The heart of the technique is to distribute sensitive data among multiple servers. As an MIT article on the topic explained, to store the number x, the system might send a random number, r, to one server, and x-r to another. Neither server would be able to calculate x on its own. But together, they could "still perform useful operations." If a cybercriminal wanted to figure out what x was, he or she would need to break into both servers — or as many servers as were involved. As servers are added to the setup, the cryptography approach becomes more complicated.

Association studies involve a massive table — or matrix — that maps the genomes in the database against the locations of genetic variations. These variations typically number about a million, requiring a million-by-million matrix, making security a complicated affair and the research effort time-consuming.

But Cho, Berger and Wu have developed techniques to simplify the security calculations and speed up the processing of their system. Based on those techniques, the system accurately reproduced three published genome-wide association studies involving up to 23,000 individual genomes. The approach could feasibly scale to a million individuals, they predict.

"As biomedical researchers, we're frustrated by the lack of data and by the access-controlled repositories," said Berger, a professor of math. "We anticipate a future with a landscape of massively distributed genomic data, where private individuals take ownership of their own personal genomes, and institutes as well as hospitals build their own private genomic databases. Our work provides a roadmap for pooling together this vast amount of genomic data to enable scientific progress."

The paper is available behind a registration wall at Nature Biotechnology.

About the Author

Dian Schaffhauser is a former senior contributing editor for 1105 Media's education publications THE Journal, Campus Technology and Spaces4Learning.

Featured

  • networked node grid featuring illuminated connectors

    Infrastructure Accounts for More than Half of Worldwide AI Spending

    Gartner forecasts worldwide AI spending will reach $2.67 trillion in 2026, up 49.5% from 2025, with AI infrastructure accounting for nearly $1.5 trillion, or about 56% of the total. For every dollar Gartner expects to be spent on generative AI models, more than $52 will be spent on infrastructure.

  • CIS Sandbox Tutor Eli Blouin at Bentley University imagines a future technology learning partner that supports curiosity, critical thinking, feedback, and personalized learning.

    Student Voices: Technology as a Future Learning Partner

    A data analytics and marketing major at Bentley University and a distinguished lecturer at the university explore the future of "technology learning partners" — technologies that participate in the learning process by asking questions and giving feedback, not just information search results.

  • young man hand working on a laptop, bokeh blurred at night

    Redesigned Copilot Will Build Apps and Work on Its Own

    Microsoft recently announced that it's redesigning Copilot, the company's all-encompassing AI assistant, so users can hand off complex tasks, build tools by describing what they need and assign work to an agent that keeps going after they log off.

  • Abstract glowing network nodes forming conversation bubbles

    OpenAI Introduces Dots, AI Agents that Continue Working Between User Conversations

    OpenAI has announced Dots, GPT-6 Astra-powered agents that run on their own cloud computers, using connected applications to carry on with assignments between conversations and incorporating users' feedback as they go.