Total $30.7 million NIH investment expands USC-led AI effort to decipher Alzheimer’s disease

The National Institutes of Health has launched new support for Artificial Intelligence for Alzheimer’s Disease (AI4AD). The new $12.6 million award to advance the next phase of the project, AI4AD2, brings the total investment in AI4AD to $30.7 million. This multi-institutional initiative, led by Paul M. Thompson, Ph.D., associate director of the Mark and Mary Stevens Neuroimaging and Information Institute (Stevens INI) at the Keck School of Medicine at the University of Southern California, will develop artificial intelligence (AI) tools to uncover the biological causes of Alzheimer’s disease and related dementias, improve predictions of disease progression, and assist in the development of more accurate treatment options.

AI4AD2 unites 10 researchers and 23 collaborators from 10 institutions to pursue four interrelated research goals. The consortium will analyze large datasets including whole genome sequencing, brain imaging, cognitive testing, and other biological data to advance dementia diagnosis and treatment. The study builds on the original AI4AD initiative launched in 2020, which developed AI tools to detect Alzheimer’s disease-related patterns in brain scans and showed how machine learning can link imaging findings to underlying genetic risks.

“As we get older, our brains decline,” Thompson says. “However, each of us has a unique combination of degenerative processes going on in our brains. We may have a mix of brain changes typical of Alzheimer’s disease, vascular disease, and Parkinson’s disease, all of which progress at different rates. This combination of pathology makes dementia difficult to treat. AI4AD2 Using this, we will launch a genome-based drug discovery program to help researchers identify new drugs that target specific types of dementia, including rarer types of dementia. ”

One of the first goals of AI4AD2 is to go beyond broad diagnostic labels and identify meaningful subtypes of Alzheimer’s disease and related dementias. Rather than grouping all patients together, the project uses AI to classify individuals based on patterns in brain scans, cognition, neuropathology, and genetic data. Improving the subtyping of dementia will improve the design of clinical trials by allowing scientists to better match treatments to patients who are most likely to benefit. These molecular subtypes are becoming increasingly important as new treatments target amyloid, tau, vascular damage, and inflammation, which affect all patients to varying degrees.

AI4AD2 will also develop new “genomic language models,” a type of AI inspired by the same broad technologies used in language-based artificial intelligence systems. Instead of analyzing words, these models analyze genome sequences to identify combinations of DNA changes associated with Alzheimer’s disease, disease progression, and key biomarkers. This project will train and evaluate these techniques using data from more than 58,000 participants across 57 cohorts. In practice, this involves teaching AI to search vast genetic datasets for patterns that cannot be identified using traditional methods. The goal is to uncover novel gene- and protein-related changes that may help promote neurodegeneration and link them to measurable changes in the brain and behavior. Previous AI4AD research showed that an AI model could identify Alzheimer’s disease-related features on brain scans with more than 90% accuracy by learning from 80,000 brain scans, demonstrating the potential of combining imaging, genomics, and machine learning at scale.

Another key focus of AI4AD2 is making sure these AI tools work well for people around the world. Many existing biomedical datasets focus on people of European descent, limiting their ability to identify risk factors that differentially impact other groups. AI4AD2 adapts its disease classification, subtyping, and prognostic tools to global and multi-ancestry cohorts, including datasets from African, Indian, Korean, and US populations. The project will identify how ancestral, social, and environmental factors influence Alzheimer’s disease risk and progression, with the aim of developing more accurate predictive models.

“The power of artificial intelligence is determined by the data and scientific questions behind it,” said Dr. Arthur W. Toga, director of the USC Mark and Mary Stevens Institute for Neuroimaging and Informatics. “This renewal enables our team and collaborators to integrate imaging, genomics, and other biomarkers to better capture the complexity of Alzheimer’s disease, allowing us to study it at a scale previously out of reach. This is an important step toward more accurate, comprehensive, and actionable brain health research.”

The project’s fourth goal focuses on discovering treatments using an approach known as genome-based drug discovery. Using a system called PreSiBO, an AI-based drug discovery tool developed through the original AI4AD effort, researchers will identify subtype-specific therapeutic targets and assess whether existing drugs can be repurposed for patients with specific Alzheimer’s disease-related biological profiles. This project will develop AI tools to detect multiple affected molecular pathways and identify specific drug treatments that target these specific disease mechanisms.

Stevens INI will continue to serve as the primary hub for this effort. AI4AD2 is designed as a highly collaborative agreement, with USC serving as the lead site and partner institutions contributing expertise in neuroimaging, genomics, statistics, machine learning, cognitive science, and drug discovery. The team will share software and tools through public repositories and scientific workshops so that researchers around the world can use and build on the project’s methods.

For families affected by Alzheimer’s disease, long-term goals are clear. The goal is to develop more accurate tools to better differentiate between different types of dementia and identify the most appropriate treatment for individual patients. By combining large-scale data and advanced AI, AI4AD2 aims to bring personalized medicine for one of the world’s most devastating neurological diseases closer to reality.