Protein folding modeling with neural cellular automata using the face-centered cubic model

Abstract

We have modeled the protein folding process with cellular automata using the Face-Centered Cubic lattice model. An artificial neural network implements a cellular automaton-like scheme that defines the moves of each of the amino acids of the protein chain and through several time iterations until a folded protein is obtained. Differential Evolution was used to evolve these neural cellular automata, which take the information for defining the folding process from the energy space considered in the lattice model. Different proteins were used for testing the process, comparing the results of the folded structures against other methods of direct prediction of the final folded conformation.