Multi-Objective Approach for Protein Structure Prediction

This work proposes to optimize Protein Structure Prediction (PSP) using multi-objective ab initio approach. This paper addresses an application of modified NSGA-II (MNSGA-II) by incorporating controlled elitism and Dynamic Crowding Distance (DCD) strategies in NSGA-II for PSP by minimizing free Potential Energy (PE) and minimizing Solvent Accessible Surface area (SAS). In this model, a trigonometric representation is used to compute backbone and side-chain torsion angles of protein atoms. Free energy is calculated using Chemistry at HARvard Macromolecular Mechanics (CHARMm -22). SAS is calculated using dssp program. Both objectives together evaluate the structures of protein conformations. The evolution of protein conformations is directed by optimization of protein energy and surface area contributions using MNSGA-II. To validate the Pareto-front obtained using MNSGA-II, reference Pareto-front is generated using multiple runs of single objective optimization (RGA) with weighted sum of objectives. TOPSIS technique is applied on obtained non-dominated solutions to determine Best Compromise Solution (BCS). Result of MNSGA-II is compared with NSGA-II. The proposed model is validated with Met-enkephalin, a benchmark protein, obtaining very promising results.