(594a) Invited Talk: Reframing of Protein Engineering in Light of Recent ML Advances

Advances in machine learning (ML) and artificial intelligence (AI) have opened new vistas in leveraging computations for protein structure prediction and design. Rapid progress has so far made possible sequence-to-structure prediction, de novo protein target structure matching with diffusion models, and the use of protein language models to learn the underlying grammar of protein sequence landscape. Such advances are driving transformative discoveries across various fields by enabling the design of new therapeutic proteins, engineered enzymes, and biomaterials. In this talk we will provide an overview of many of these developments. In addition, we will discuss forays in this space by our group for quantifying the binding affinity of SARS-CoV-2 RBD variants vs. the mammalian ACE2 receptor binding. The methodology combines biophysical information as features gleaned through molecular dynamic simulations with a neural network approach. On a separate challenge, we introduced CatPred, a deep learning framework that utilizes protein language models to provide predictions of in vitro enzyme kinetic parameters with uncertainty quantification. By integrating function prediction tools, the hope is that powerful generative AI pipelines for predicting protein function and structure will become achievable in the near future.