A genetic algorithm with conformational memories for structure prediction of polypeptides.

Area of Science:

• Computational biology
• Biophysics
• Structural bioinformatics

Background:

• Predicting peptide three-dimensional (3D) structures from amino acid sequences is crucial for understanding their function.
• Existing methods often face challenges in efficiently exploring the vast conformational space and locating the global energy minimum (GEM).

Purpose of the Study:

• To develop and validate an iterative hybrid algorithm (HA) for accurate peptide 3D structure prediction.
• To enhance the efficiency of finding the putative global energy minimum (GEM) by incorporating conformational memories.

Main Methods:

• The hybrid algorithm (HA) integrates a modified genetic algorithm (GA) with a local optimizer.
• HA employs a two-phase iterative approach: GA runs exploring conformational space, followed by utilizing 'conformational memories' to refine the search.
• Conformational memories are used to reduce the search space in subsequent iterations, avoiding conformational barriers.

Main Results:

• Successfully predicted the putative GEM for Met- and Leu-enkephalin.
• Obtained valuable 3D structure information for polyglycine 8mer and a 16-residue (AAQAA)(3)Y peptide.
• Demonstrated fewer fitness function evaluations compared to other heuristic methods for locating putative GEMs.

Conclusions:

• The developed HA offers an efficient and effective approach for peptide 3D structure prediction.
• The incorporation of conformational memories significantly speeds up and refines the localization of the global energy minimum.
• This study highlights the potential of genetic algorithms for high-level polypeptide secondary structure prediction.