Metaheuristics for Protein Structure Prediction: Review and Empirical Analysis

Metaheuristics have been employed in solving several optimization and NP-hard problems owing to their ability to avoid local optima entrapment, flexibility and robustness, simplicity, and reasonable computational time. In this work, an empirical analysis is carried out on the application of metaheuristics in solving the protein structure prediction (PSP) problem. The PSP problem entails predicting a three-dimensional (3D) structure of proteins based on the amino acid sequences. This study employs the Ab Initio method, which relies on the thermodynamic properties of the proteins, such as the minimal free energy levels, bond angles, and torsional angles, in determining the protein structure. Five real protein sequences, namely 1CB3, 1BXL, 2H3S, 1TZ4, and 1CRN from the Protein Data Bank (PDB), and 11 metaheuristics including DE, GA, PSO, DSO, ACO, and HS were employed in the empirical analysis. Moreover, extensive Monte Carlo simulations were carried out to generate the results of the empirical analysis and these results were statistically tested using the Friedman test.