Refinement of Protein Structures with a Memetic Algorithm. Examples with SARS-CoV-2 Proteins

Considerable progress has recently been made in protein structure prediction with deep learning methods. However, the structures predicted with these methods often require a subsequent refinement process to correct imperfections in the prediction. This paper uses and compares different possibilities for this refinement, including our memetic algorithm, which combines differential evolution with a state-of-the-art refinement method from the Rosetta environment. The comparison and analysis are performed with selected proteins from the SARS-CoV-2 virus.