Across the globe, scientists are on a quest to find the “unknown known” – a term that Matthew Might, Hugh Kaul Endowed Chair and director of the Hugh Kaul Precision Medicine Institute at the UAB Heersink School of Medicine, uses to describe the phenomenon of scientific facts hiding in plain sight.
It turns out there are two kinds of hidden facts:
Many of these facts hiding in the unknown known have direct implications for diagnosing or treating individual patients – or for new approaches to entire diseases.
In biomedicine, massive data sets exist independently; but an effective method has not yet been deployed to connect the dots from disease to data, to treatment, especially for rare diseases. According to the National Human Genome Research Institute, rare diseases affect between 25 million and 30 million Americans each year.
“There are billions of facts sitting around,” Might said. “We know there is new data on diseases and treatments, theoretically; but no one has made the deductions yet.”
A cutting-edge consortium project called the Biomedical Data Translator aims to make those missing deductions through artificial intelligence. It is funded by the National Center for Advancing Translational Sciences (NCATS) at the National Institutes of Health.
The project is transferring all scientific data in the world into the Translator and developing an automated reasoning agent, a “virtual brain” that will reveal the unknown knowns of diseases and associated therapeutics. It is the first of its kind. To date, no comprehensive biomedical knowledge source of this size has been built.
“The aim of the consortium is to render all available biomedical knowledge as interconnected knowledge graphs and make them accessible through standardized APIs, or application programming interfaces,” Might said.
A partnership between Beshenich Muir and Associates and the Kaul Precision Medicine Institute has been selected by NCATS as recipient of the Biomedical Data Translator User Interface Program Award, a federal contract that will support the development phase of the Translator user interface.
The funding, $1.3 million per year for three years, will allow PMI and Beshenich Muir to hire several full-time employees to develop the user interface, in conjunction with the consortium’s development of the core Translator system and knowledge base.
Building on mediKanren
Because of its previous work on mediKanren, a reasoning engine for biomedical knowledge created by Might, PMI is uniquely situated to contribute its expertise to the Translator project. Like mediKanren, the Translator project has one major ambition: to accelerate biomedical research. To do this, researchers and technologists will structure all scientific data, inputting it into the Translator. It will then be converted into accessible biomedical knowledge. Then the Translator can make deductions on its own using automated reasoning.
Several data types will be considered for the Translator, including disease data, symptoms, biomarkers and genes, clinical trial data, diagnostics and much more. Over the course of 20 years, milestones will mark progress: compiling data into the Translator, applying reasoning over the data and exposing the knowledge for scientists and physicians. Once complete, the Translator should answer researchers’ unresolved questions, such as predicting a treatment for a disease or learning how gene variants respond to drug treatments.
To develop the Biomedical Data Translator, several teams have been or will be deployed to integrate extensive, currently available medical research data. NCATS has issued awards to project teams of experts from leading universities and research institutions.
Might explained that, generally speaking, three types of teams work on the Translator: knowledge teams, reasoning teams and, now, an interface team.
Toward a data-driven future
Might said the beta version of the Translator interface will not be immediately available to the general public, but “our hope is that, one day soon, even this beta version will begin helping translational scientists and physicians accelerate their work.”
PMI hopes to pull in scientists from varying fields, especially from UAB, to understand the types of questions they are trying to answer. “If we know what scientists are trying to deduce, it will inform how the user interface is built,” Might said.
Leveraging data with artificial intelligence tools will shape the future of medicine.
“I’m excited about this opportunity – I have seen how powerful this is in the context of our own patients,” Might said. “In some sense, I can imagine what will happen, but I am excited about the unknown known, too – to sit back and watch what researchers will do when they have access to all this knowledge.”
Partners on the Translator UI Award include Elena Glassman, associate professor of computer science at Harvard University, and experts from the University of Colorado, Melissa Haendel and Casey Greene.
This story originally appeared on the UAB News website.