17-year-old builds AI to decode protein structures, wins $90,000

Jerry Xu's AI model is designed to compare the structures of proteins and amino acids rapidly. The structure of proteins gives important clues to how these molecules function. (Images: istock and Society for Science) Imagine, what if a protein's complex 3-D structure could be converted into a simple string of numbers—and that string could help scientists understand what the protein does? That’s the premise of an innovative AI model developed by 17-year-old Jerry Xu of Lexington, Massachusetts, who earned $90,000 in the Regeneron Science Talent Search for his research, states an online report by the Society for Science. Scroll down to read the details.Converting protein structures to numbersAccording to the online report, Xu’s project, “A Unified Protein Embedding Model With Local and Global Structural Sensitivity,” is in the area of computational biology and bioinformatics. His AI model is designed to compare the structures of proteins and amino acids rapidly. The structure of proteins gives important clues to how these molecules function. But it can take huge amounts of computing power to find and compare all the details in those structures. Another problem facing scientists is th...