George Mason University,
Keywords: AI, biomedical application, machine learning
Summary:Rapid advances in genome sequencing held the promise that we would soon know what genes do, but this has proven more challenging than anticipated. We are still not able to predict how sequence variation in a protein translates to phenotypic alterations. Even AlphaFold2, with its promise of having solved the decades-old protein structure prediction problem, cannot capture fundamental knowledge about protein structure plasticity and its central role in protein function. In this talk I will overview work in my laboratory on AI frameworks that link protein sequence variation to dynamics-governed dysfunction. I will also describe some of our deep generative frameworks building on variational autoencoders or generative adversarial networks and advance the capability of these frameworks to map protein structure space. Finally, I will present some promising directions that showcase the power of deep generative frameworks to link chemical and biological space in small, drug-like molecules and allow us to generate small molecules in-silico with control.