Structure prediction on large protein using the combination of knowledge based and physics based approaches method validation on cholesterol esterase

Abstract

The objective of this study was to predict the structure of a large protein using a combined approach of knowledge-based comparative modeling and physics-based Molecular Dynamics (MD) simulation applied to the enzyme cholesterol esterase. The core region of the enzyme was modelled using information from homologous known protein structures whereby leaving the end-terminal regions (the nonhomologous regions) to fold via MD simulation. Currently, there is yet a reported study where one begins with a knowledge-based model of the core region of a protein and allowing the remaining end terminal regions to fold via MD simulation. The method was categorized into three parts; ~ the development of the core region of the protein, the development of the complete protein structure and the MD refinement simulation. Three models were tested, CECRL-87, CETHG-45 and CEn~M-14, with each originating from different core regions developed at three different cutoff values of sequence identity; more than 70% (%id > 70%), less than 60% but more than 30% (30% < %id < 60%) and less than 20% (%id < 20%), respectively. The remaining residues were later added using MD simulation which then followed by 20 ns of MD refinement. It was shown that the use of different starting core regions did not significantly contribute towards correct structure predictions of large proteins. Furthermore, the use of restraint of the core region would only deteriorate the model as observed in CETHG-45.