CMU Researchers Develop Tools To Target Brain Disorders

Researchers from Carnegie Mellon University's Kenneth C. Griffin School of Computer Science and theUniversity of Pittsburghare developing molecular tools to better understand specific brain cells and, ultimately, treat brain disorders, such as depression, in a more targeted way. The tools being developed aren't necessarily ones you can use to poke or prod — they are genes or proteins targeted to elicit specific outcomes and avoid unpleasant side effects. "The main problem in studying the brain is its complexity," saidAndreas Pfenning, an associate professor in theRay and Stephanie Lane Computational Biology Department. "There are so many different subtypes of cells in the brain and neurons performing so many different functions that when we're trying to study the brain or treat it, we often see sledgehammer approaches. For example, if we modulate dopamine, then that influences many of the systems in the brain that depend on dopamine, not just mood, and a lot of people have a lot of side effects." This work is being done in partnership with Pitt and is supported by a multiyear grant from theNational Institutes of Health's Armamentarium for Precision Brain Cell Access initiative, which funds research to develop molecular genetic reagents to target specific neural circuits. From previous research, scientists can now identify many distinct types of brain cells, but they still lack reliable tools to reach and manipulate most of them with precision. When precise targeting sequences exist, they're often slow and costly to test individually, may behave differently across brain regions or experimental conditions, and can activate the wrong cells or too few of the intended cells to be useful. Even when effective tools are developed, the supporting data, protocols and materials are often scattered across publications and laboratories, making results difficult to reproduce or apply. Together, these limitations create a major gap between mapping the brain's cellular diversity and understanding how specific cells and circuits contribute to brain function and disease. This funded phase of the Armamentarium initiative seeks to turn rapidly expanding maps of brain-cell types into a practical toolkit for precisely reaching and manipulating those cells. Its desired outcome is not a single invention, but a scalable pipeline that combines AI-guided design, multiplexed screening, spatial measurement, functional validation and open distribution to potentially give neuroscience a powerful new level of precision for investigating brain circuits and their roles in health and disease. In this research partnership, Pfenning's lab will focus on integrating AI methods with newly developed wet lab genomic technologies — the laboratory experiments used to study and manipulate genetic material — to create these molecular tools. Meanwhile, research partners includingWilliam Stauffer, an assistant professor in neurobiology at Pitt, will validate the new tools and further develop them for use in the laboratory. All this work will lead to a more comprehensive knowledge of the brain. "The brain is made up of these different types of neurons," Pfenning said. "And if we're going to build a better basic understanding of how the brain works and how it evolved, we need to have tools that can study the function of these different subtypes of cells." Instead of relying on slow trial and error, the researchers will use AI to design customized molecular tools, like specialized nucleic acids, that target specific brain cells. New technology developed at CMU enables simultaneously testing dozens of designs, meaning the team can rapidly refine these tools and share them with the scientific community. This method, still in its early stages, could also open the door to more targeted gene therapies for rare diseases. "It's essentially an iterative loop between AI and experimentation," Pfenning said. "In this case, we plan to develop the genomic technology necessary to collect the right data to refine our computational models, and then this data will help us optimize the tools for specific cell types." The researchers have already begun their work by training AI models and, in parallel, growing the cells they'll use for the first round of experiments. For more information on Pfenning's work,visit his lab's website.