WashU Medicine has been awarded nearly $800,000 by the National Artificial Intelligence Research Resource (NAIRR) Pilot program, an initiative of the National Science Foundation (NSF), and Microsoft to develop AI tools aimed at improving biomedical research efficiency, with an initial focus on Alzheimer’s disease.
The grant will fund the Consortium for Biomedical Research and AI in Neurodegeneration (C-BRAIN), led by director Randall Bateman, MD, the Charles F & Joanne Knight Distinguished Professor of Neurology, and associate director Eric Landsness, MD, PhD, an assistant professor of neurology, in their efforts to build an “AI Scientist” system to boost the efficiency and effectiveness of Alzheimer’s disease research. The researchers are focused on Alzheimer’s disease because despite tremendous investments over many decades, 99% of drug candidates for this disease fail to show any benefit in clinical trials.
The goal of the AI Scientist system is to gather and make sense of vital knowledge about Alzheimer’s disease that is currently fragmented across millions of papers and enormous and complex datasets, some of which are unpublished. The volume of information is only increasing, demanding newer and more efficient methods for analysis.
“A fundamental challenge in creating a useful AI ‘teammate’ is developing a system that can perform tasks that a researcher cannot do, such as identifying meaningful connections across massive amounts of data in minutes,” said Adith Boloor, PhD, a staff AI scientist with the WashU Digital Intelligence and Innovation (DI2) Accelerator and the grant’s technical lead. “It also can help researchers with tedious, time-consuming tasks like the curation of unstructured data. An effective AI collaborator must augment human expertise by enabling large-scale data synthesis, while also automating routine and time-consuming work — both executed with transparency and reliability.”
The results of the project will be shared publicly through publications, code and data resources.
“This project will deliver substantial understanding on the science of doing science, in addition to advancing knowledge for the development of Alzheimer’s drugs and other discoveries,” said Landsness.
The AI Scientist will include three interrelated products:
- An AI literature synthesizer that will ingest existing published research, trial reports, and other data and search it for patterns. From these patterns, it will propose various hypotheses related to neurodegeneration.
- An AI data analyzer to take those proposed hypotheses and test them against existing published results and datasets.
- An AI reviewer to provide peer-review style critiques of the hypotheses, experimental results and interpretations.
Human scientists will evaluate and refine each product to ensure accuracy. Together, the three products will create an “open standard for AI-scientist systems that enables others to evaluate and contrast their workflows within a transparent, comprehensive discovery framework,” said Bateman.
Collaborators on the grant include members of WashU Medicine Neurology, the WashU Research and Development Office, the WashU DI2 Accelerator and several external contributors.