The rise of therapies that target tumors based on their genetic profile is revolutionizing how we treat cancer. It’s also creating a data overload.
Around 17 million people are diagnosed with cancer each year, and hundreds of cancer studies are published each day. The challenge for doctors is narrowing that down to one patient and their treatment.
“There’s so much data around. It really becomes a complex question,” said Dr. Sue Mockus, director of product innovation and strategic commercialization at the Jackson Laboratory (Jax).
Microsoft and Jax created a system, called Clinical Knowledgebase, that uses natural language processing to find clinically relevant needles in the haystack of published studies.
In order to match a tumor’s genetic profile to the right treatment, hospitals often employ a tumor board, which is comprised of a group of experts from different specialties who can assess the available treatments. Clinical Knowledgebase aims to supercharge oncologists with what amounts to a search engine that can return all of the relevant sections from thousands of studies in a few seconds.
“Sequencing has become really cheap and affordable. Very soon we can actually afford to sequence basically every cancer patient,” said Hoifung Poon, director of precision health natural language processing at Microsoft. The challenge, he said, will be acting on that genetic information efficiently.
The goal is to make the most of human specialists and computers. Machines are great at scaling — hence their role in filtering out studies — while humans are best at applying specific knowledge to find the most promising treatment for each patient.
“What we have been focusing on is trying to leverage the sweet spot of human-computer symbiosis,” Poon said.
The effort is part of Project Hanover, an initiative launched by Microsoft in 2016 to “solve” cancer. More than 70,000 users across 149 countries have taken advantage of Jax’s platform already.
The Jackson Laboratory is a 90-year-old nonprofit research institution in Bar Harbor, Maine, that is known as the repository for thousands of genetic strains of mice used in clinical research.
Microsoft’s growing list of partnerships with healthcare institutions includes Humana, Adaptive Biotechnologies, Providence St. Joseph Health, UCLA Health and University of Pittsburgh Medical Center. In a separate effort, Microsoft created a chatbot that matches patients to clinical trials. Among tech giants, Google parent Alphabet has arguably been the most active in healthcare research efforts, notably using AI to diagnose eye disease and breast cancer.