A Duke University research team is using NCShare resources to speed up the discovery and development of disease treatments with AI. Rohit Singh, assistant professor in the departments of biostatistics & bioinformatics and biochemistry & cell biology at Duke, leads the project, which focuses on enabling new biotechnology and therapeutic advances by designing more effective genes and proteins.
“One of the things that was made possible by the NCShare GPUs is to dramatically enhance the power of our generative AI software that can take any natural protein and shrink it,” Singh said. “What it does under the hood is almost like how large language models are good at summarizing long text. It uses the same ideas to summarize a large protein or gene into something smaller. That matters because even though proteins are already microscopic, the cell is also very small. Proteins that are smaller can be easier to handle in biotechnological applications and easier to deliver therapeutically.”
Singh calls this protein-shrinking program Raygun as an homage to the movie “Honey, I Shrunk the Kids.” He is currently working on the next version of Raygun. When complete, Raygun 2 will scale up the project’s computing power by a factor of 10, allowing for many new discoveries. Singh said NCShare allows his team to be innovative and compete with large labs across the country by providing high-caliber resources efficiently.
“I think NCShare is a great resource because it lowers the cost of adoption,” Singh said. “One of the big challenges in GPU computing is not knowing what hardware to get or how to set everything up. Here, everything is set up for you, so you can log in and immediately be productive, which is immensely useful. I would argue that this is potentially even more useful for smaller institutions trying to figure out how to get their trainees into AI or give researchers access to AI because it removes the burden of setting up infrastructure and making purchasing decisions.”
In the future, Singh said he hopes NCShare can democratize GPU computing even more across North Carolina.
“We can publicize this to the community, the legislature, and the executive branch so they see how these investments, even if small in the grand scheme of things, are immensely high-leverage and make large achievements possible,” Singh said.