3D structure predictive performance of de novo modeling based on molecular dynamics simulation for polypeptides with experimentally solved structures

Abstract

Introduction: Proteins are made up of building blocks called amino acids. The wide variety of proteins' functions, these vital constituents of living cells, is due to their specific three-dimensional structure originating from the appropriate folding of amino acid sequence. Protein structure determination has long been of great interest and challenge in the field of molecular design. Aims: In the current study, de novo modeling was employed to assess its predictive performance for structurally-solved polypeptides . Materials and methods: Polypeptides with 20 to 150 amino acids were extracted from a protein data bank (PDB) and their amino acid sequences were retrieved. In the next step, molecular dynamics (MD) simulation was performed for 50 ns in two environments including vacuum and solvated systems. Following the MD simulation, the stability of the whole system in terms of energy and conformational changes was investigated. Then, conformational clustering was conducted, and the structures with the lowest root-mean-square deviation (RMSD) to the centroids of the highly-populated clusters were selected. Moreover, the models with the lowest energy during the simulation time were opted. Then, the secondary structure contents were assigned for the obtained models using DSSP (defined secondary structure of proteins), PCASSO (protein c-alpha secondary structure output), and STRIDE (structural identification) algorithms. Results: Analyses revealed no meaningful relationship between the secondary structure elements of MD- and X-ray based models. The current MD-based de novo method, predicts the three-dimensional (3D) structures for the given polypeptides differently than experimental methodologies. Conclusion: The current study can provide a better insight into the capability and limitations of de novo molecular modeling of proteins.