Linna An

A general computational method to build small molecule-protein interfaces from scratch

would be widely useful for many challenges, including ligand binder, sensor design, and

enzyme engineering. Yet, designing proteins to bind small molecules is extremely challenging

because the problem requires a highly accurate sampling in a gigantic sampling space. In

this talk, I will go through how we approach the small molecule binding problem through

building machine learning-based protein design methods and successfully demonstrating

their robustness through designing binders to various ligands, including polar and flexible

ones. Using the designed binders, I further developed a general method to turn binders into

sensors through building chemically induced oligomerization systems. I will also project how

our methods can revolutionize enzyme engineering by significantly increasing our screening

capability and enabling large structural manipulations that were impossible before on enzyme

redesign.