Salp-J Colony Optimization-based advanced hybrid ensemble deep predictor with LSTM for protein structure prediction

Area of Science:

• Computational biology
• Bioinformatics
• Structural biology

Background:

• Protein structure prediction (PSP) is crucial for understanding protein function.
• Accurate protein shape determination is vital for biological research.
• Protein secondary structure prediction (PSSP) is foundational for 3D PSP.

Purpose of the Study:

• To develop an advanced hybrid ensemble deep predictor for accurate protein structure prediction.
• To improve the performance of protein structure prediction using enhanced feature extraction.
• To validate the proposed method's efficacy using benchmark datasets.

Main Methods:

• Utilized Long-Short Term Memory (LSTM) for deep learning-based prediction.
• Integrated Salp-J Colony Optimization for enhanced feature extraction.
• Combined exploration behavior of Ulmaris, virus colony immunity, and salp teamwork for optimization.

Main Results:

• Achieved 99.1% accuracy, 99.5% sensitivity, 98.85% specificity, and 0.9% error on CullPDB dataset (80% training).
• The proposed method outperformed Protein Net (97.27% accuracy, 2.7% error).
• Demonstrated significant improvements in predicting protein structures.

Conclusions:

• The advanced hybrid ensemble deep predictor with Salp-J Colony Optimization offers a highly accurate approach to PSP.
• The method effectively enhances feature extraction for improved prediction accuracy.
• This work contributes a robust tool for computational biology and drug discovery.