Protein subnuclear localization based on a new effective representation and intelligent kernel linear discriminant...

Area of Science:

• Bioinformatics
• Computational Biology
• Proteomics

Background:

• Protein subnuclear localization is crucial for understanding cellular functions.
• Existing prediction methods often rely on computationally expensive fusion of multiple feature representations.
• There is a need for accurate and efficient methods for predicting protein subnuclear localization.

Purpose of the Study:

• To develop a novel, computationally efficient method for predicting protein subnuclear localization.
• To introduce a new single feature representation, Correlation Position-Specific Scoring Matrix (CoPSSM).
• To propose an optimized algorithm for parameter selection in Kernel Linear Discriminant Analysis (KLDA).

Main Methods:

• Developed Correlation Position-Specific Scoring Matrix (CoPSSM) as a novel protein feature representation based on Position-Specific Scoring Matrix (PSSM).
• Proposed a dichotomous greedy genetic algorithm (DGGA) to optimize the kernel bandwidth parameter of Kernel Linear Discriminant Analysis (KLDA).
• Utilized Jackknife and independent tests with KNN classifier on public datasets for validation.

Main Results:

• The proposed single-feature representation (CoPSSM) achieved higher Overall Success Rates (OSR) than existing multi-feature fusion methods.
• The novel method reached prediction accuracies of 87.444% and 90.3361% on two standard datasets.
• The algorithm demonstrated considerable prediction accuracy on additional datasets, confirming its generalization ability.

Conclusions:

• The proposed method combining CoPSSM and DGGA-optimized KLDA offers a highly accurate and efficient approach for protein subnuclear localization prediction.
• Single, well-designed feature representations can outperform complex fusion strategies in terms of both accuracy and computational cost.
• This work provides a valuable tool for advancing research in proteomics and cellular biology.