Ab initio protein fold prediction using evolutionary algorithms: influence of design and control parameters on...

Area of Science:

• Computational Biology
• Biophysics
• Structural Bioinformatics

Background:

• Accurate prediction of protein 3D structure from primary sequence is a fundamental challenge.
• Evolutionary algorithms (EAs) show theoretical promise for navigating the complex energy landscape of protein folding.
• Previous studies have been limited by insufficient optimization of EA design and control parameters.

Purpose of the Study:

• To comprehensively investigate EA design and control parameters for ab initio protein fold prediction.
• To identify optimal EA configurations for accurate and efficient protein structure determination.
• To assess the transferability of optimized EA parameters to real protein systems.

Main Methods:

• A full atomistic protein model was employed for ab initio protein fold prediction.
• Twelve distinct EA designs were evaluated, varying encoding, crossover, and replacement strategies.
• Optimal control parameter settings were determined using a 15-residue polyalanine molecule.
• Parameter scaling was analyzed with increasing polyalanine size.
• The optimized steady-state EA design was applied to met-enkephalin.

Main Results:

• Real encoding and multipoint crossover were identified as superior design choices.
• Both generational and steady-state replacement strategies demonstrated effectiveness.
• Optimal control parameter settings were established and their scaling with molecule size was quantified.
• The optimized steady-state EA design showed promising performance on met-enkephalin, suggesting transferability.

Conclusions:

• Careful selection of EA design and control parameters is essential for successful ab initio protein structure prediction.
• Optimized EAs, particularly the steady-state design with real encoding and multipoint crossover, are competitive with other methods.
• This work provides a framework for enhancing EA performance in computational protein structure prediction.