When Xiang Yu watches his 4-month-old daughter playing, he can’t help but be inspired. One of her teething toys is made of colorful loops of soft plastic, knotting around each other to create a unique, 3D structure — not unlike a protein. At Merck, where Yu is a scientist in the pharmacokinetics, pharmacodynamics, and drug metabolism group, his job is to untangle those knots, helping drug developers understand the way drug candidates are metabolized.
Since joining Merck in 2015, Yu has worked to identify promising drug candidates as quickly as possible. After getting his Ph.D. in analytical biochemistry at Boston University and a postdoc at Northwestern, he realized that biochemistry couldn’t provide a solution to all his questions — so he sought out new strategies in computer science, getting a master’s while working in the lab and raising his infant son. “I developed a technique with the baby on my left arm, and coding with my right,” Yu laughed.
Today, Yu is the data science lead in his group, heading up development of an artificial intelligence platform he calls Merckoid, which aims to support decision-making in preclinical drug development. “When you start with thousands or millions of molecules, you have to make layers of decisions where you triage to just this one,” said Yu. “It would be really nice to have something like Alexa — like ‘Hey, Alexa,’ or ‘Hey, Merckoid’ — with an AI system that can behave like Merck scientists.” The system is still in its early stages, but “if we can’t fully automate it,” said Yu, “at least it can augment our human performance.”
— Katie Palmer
