Scientists use AI to identify likely drug targets in search for Alzheimer's cure

Image
Rui Chang

Rui Chang, an associate professor of neurology, is leading a team at the College of Medicine – Tucson that uses artificial intelligence to trace the molecular path of the development of Alzheimer's disease.

Researchers at the University of Arizona College of Medicine – Tucson, along with collaborators at Harvard University, harnessed the power of artificial intelligence to identify causes of Alzheimer's disease and potential drug targets by looking deep into the human brain to map the molecular changes that healthy neurons undergo as the disease progresses. The results are published in Nature Communications Biology.

One of medicine's most perplexing problems is Alzheimer's disease, a neurodegenerative disorder that causes dementia, memory loss, personality changes and other irreversible symptoms. Drugs can treat symptoms of the disease, but finding a cure has been challenging, possibly because the cause of Alzheimer's disease is unclear.

"There are multiple pathways involved in Alzheimer's disease," Rui Chang, associate professor of neurology, said of the sequence of events that occur in cells to trigger changes in the body. "This is the first study showing that the AI and big data-driven approach could open the door to develop treatment for Alzheimer's by targeting new pathways or combinations of pathways."

With tissue samples from more than 2,000 Alzheimer's brains taken from a national database, Chang's AI algorithm drew from a deep well of information about genetic and molecular processes, returning a computational network model of the human brain. His team can now see maps of whole-genome genes that work together and can track the sequential changes in these genes' relationships as Alzheimer's develops, providing clues to the disease's origins and tracing the molecular path from health to disease.

Chang likens the path from health to Alzheimer's to a watercourse in which the buildup of telltale amyloid plaques and tau tangles, abnormal structures in the Alzheimer's brain, occur "downstream" in response to problems occurring "upstream."

"Amyloid plaques and tau tangles are downstream effects of a series of genetic mutations in upstream pathways that induce Alzheimer's. It's very doubtful that targeting these abnormal structures directly will be effective," Chang said.

Drugs that clear away and halt production of plaques and tangles have failed in clinical trials, pointing to the likelihood that they are not a cause of Alzheimer's but rather a consequence of earlier events.

"In my perspective, the correct way is to target the disease upstream and especially multiple targets simultaneously. Therefore, it is critical to understand the whole landscape," said Chang, who uses AI to map this landscape. "AI is a novel method that can tease apart massive data into a network model to provide a crystal-clear picture of entire upstream events, showing which upstream genes coordinately control important downstream genes. With that model, we are able to pinpoint the upstream genes that trigger amyloid plaques and tau tangles downstream. These upstream genes may be better targets for potential therapies."

Chang used AI to help identify 19 particularly interesting neuron-specific genetic points on the Alzheimer's pathway that appear to push neurons closer toward a disease state. Study collaborators at Harvard validated these genes' role in Alzheimer's development by using stem cells to create neurons in petri dishes and then deactivating the genes to see what would happen. They found that 10 of these genes affected production of plaques and tangles and could be investigated as targets for drugs to treat Alzheimer's.

"If deactivating these genes significantly changes the levels of amyloid plaques and tau tangles, then this is a validated target to develop treatments for Alzheimer's disease," Chang said.

Once gene targets are identified, the next step is to find drugs that will hit those targets. Chang used 3D computer models determine if existing molecules and drugs might fit into potential drug targets like a lock into a key.

"This is not studying one gene by one gene, it is 6,000 targets all at the same time, which will significantly accelerate drug development and discovery," Chang said.

The team virtually screened millions of Food and Drug Administration-approved, natural product and small-molecule compounds against more than 6,000 targets, narrowing the field to about 3,000 drug candidates of interest. Along with several small molecules they are investigating further, the team already has a National Institutes of Health grant enabling clinical trials on three of the compounds. They expect human trials to start soon.

"Starting from mathematics and data, I can design mathematical algorithms to lift up a massive amount of data all the way to clinical studies in patients," Chang said. "To see the compounds going into clinical trials and finally benefit patients is a fascinating journey for me."

This research was conducted using a database made available by the Accelerating Medicines Partnership Program for Alzheimer's Disease (AMP AD), a public-private partnership between the National Institutes of Health (NIH), the U.S. Food and Drug Administration (FDA), multiple biopharmaceutical and life science companies, and nonprofit organizations, to transform the current model for developing new diagnostics and treatments for Alzheimer's disease. This work was also supported by National Institutes of Health grants NIH/NIA 1R56AG062620-01, 1RF1AG057457-01, R01AG055909 and RF1NS117446, NIH/NINDS U54NS110435, and institutional funds from the University of Arizona Center for Innovation in Brain Science.

This story was originally published on the UArizona Health Sciences website.

Resources for the Media

Media Contact(s)