Artificial intelligence accelerated protein function prediction through structure-based docking

With the recent adoption of state-of-the-art machine learning methods for the determination of 3D protein structure, there is a need for high-throughput methods for the determination of previously unknown protein function. This work takes advantage of the structure-function relationship of proteins to make bio-informed functional predicitons for un/under-characterized protein sequences. In this context, we present an end-to-end workflow for the characterization of proteins within the bacteria Pseudomonas fluorescens through a series of tools such as protein structure prediction, compound screening, and molecular docking. This enables a high-throughput mechanism for initial computational predictions that can be further confirmed through experiment. The results allow for a greater understanding of each protein's role in a given genome or community-scale model.