MGM graduate student Justice Lu won first place in the Abstrax Prize for Applications in Analytical Chemistry in the 2025 Large Language Model Hackathon for Materials Science and Chemistry.
His project, “SmeLLMap: Turning smell from a mystery into a map,” used advanced computer tools to build a system that can guess how receptors in our noses respond to different smells based on their structure.
While scientists understand how our eyes turn light into images, they don’t yet have a clear map of how our noses turn chemicals into smells.
“By integrating AlphaFold-predicted models and protein large language models to build a deep learning model, the project demonstrated that structural features of receptors can be used to predict odor responses,” Lu said.
“The findings highlight the receptor binding cavity as a key determinant of chemical recognition, offering new insights into the molecular logic of smell with implications for fragrance design, and drug discovery,” he explained.
The Large Language Model Hackathon for Materials Science and Chemistry is a hybrid international hackathon connecting researchers from across the globe to explore and help solve the biggest problems in materials science and chemistry.