PROTEIN STRUCTURE PREDICTION: AN IN-DEPTH COMPARISON OF APPROACHES AND TOOLS

Abstract

Proteins play crucial roles, including biocatalysis, transportation, and receptor activity, in living organisms. Moreover, their functional efficacy is influenced by their structural properties. Determining the three-dimensional structure of a protein is crucial to comprehending its catalytic mechanism, identifying potentially beneficial mutations for industrial applications, and enhancing its properties, including stability, activity, and substrate affinity. Although X-ray crystallography, nuclear magnetic resonance (NMR), and electron microscopy are employed to ascertain protein structures, many researchers have turned to bioinformatics modeling tools because of the high cost and time demands of these techniques. For structure prediction, there are three basic methods: ab initio (de novo), homology-based, and threading-based modeling techniques. In this study, 11 modeling tools belong to different approaches were compared through modeling of various proteins; Geobacillus kaustophilus ksilan alpha-1,2-glucuronidase, Actinosynnema pretiosum bifunctional cytochrome P450/NADPH-P450 reductase, human high affinity cationic amino acid transporter 1 (SLC7A), human proton-coupled zinc antiporter (SLC30A) and Bacillus subtilis RNA polymerase sigma factor (sigY). Generated models were validated through QMEAN, QMEANDisCo, ProSA, ERRAT and PROCHECK tools. All of the studied proteins could be successfully modeled using homology modeling techniques, while some of the proteins could not be effectively modeled using threading or ab initio-based methods. YASARA generated reliable models for proteins that contain heteroatoms, such as P450 monooxygenases, because other tools exclude heteroatoms in their produced structures. Among approaches for modeling without templates, AlphaFold is a potent tool. On the other side, well-known template-based tools like YASARA, Robetta, and SWISS-MODEL have arisen. These results will help scientists choose the best protein modeling strategy and tool to guarantee high-quality structures.