Enhanced Methodology for Peptide Tertiary Structure Prediction Using GRSA and Bio-Inspired Algorithm.

Area of Science:

• Computational biology
• Bioinformatics
• Biophysics

Background:

• The Protein Folding Problem (PFP) is crucial for understanding protein function and is computationally challenging (NP-hard).
• Minimizing energy functions is key to predicting stable, biologically relevant protein structures.
• Bio-inspired algorithms offer effective solutions for complex computational problems like PFP.

Purpose of the Study:

• To develop a novel hybrid algorithm for precise protein structure prediction.
• To enhance protein structure refinement using an integrated computational framework.
• To evaluate the performance of the proposed algorithm against existing state-of-the-art methods.

Main Methods:

• Introduction of the GRSABio algorithm, integrating Jumping Spider Algorithm (JSOA) and Golden Ratio Simulated Annealing (GRSA).
• Incorporation of a Convolutional Neural Network for fragment prediction (FCNN) to create the GRSABio-FCNN methodology.
• Application of the GRSABio-FCNN approach to a dataset of 60 peptides and statistical comparison using Wilcoxon and Friedman tests.

Main Results:

• The GRSABio-FCNN approach demonstrates competitive performance against state-of-the-art methods for peptides up to 50 amino acids.
• The enhanced methodology surpasses leading PFP algorithms for predicting structures of peptides with up to 30 amino acids.
• The integrated framework shows improved energy-based structure refinement for protein prediction.

Conclusions:

• The GRSABio-FCNN represents a significant advancement in protein structure prediction, particularly for shorter peptide sequences.
• The hybrid approach effectively leverages bio-inspired strategies and deep learning for enhanced prediction accuracy.
• This study validates the potential of combining GRSABio and FCNN for tackling complex PFP challenges.