Predicting the Subcellular Localization of Proteins with Multiple Sites Based on Multiple Features Fusion

Area of Science:

• Bioinformatics
• Computational Biology
• Molecular Biology

Background:

• Protein subcellular localization is vital for function and drug discovery.
• Experimental methods are time-consuming and costly.
• Existing machine learning models often predict only a single location, but many proteins have multiple locations.

Purpose of the Study:

• To develop an improved computational method for predicting multiple subcellular locations of proteins.
• To address the limitations of existing single-location predictors.
• To enhance the accuracy of protein subcellular localization prediction for proteins with dual or multiple locations.

Main Methods:

• Feature extraction by fusing several methods to represent protein sequence information.
• Application of the multi-label k-nearest neighbors (ML-KNN) algorithm for prediction.
• Evaluation on datasets s1 (Gpos-mploc) and s2 (Virus-mPLoc).

Main Results:

• Achieved 66.7304% accuracy on dataset s1 (Gpos-mploc).
• Achieved 59.9206% accuracy on dataset s2 (Virus-mPLoc).
• Demonstrated the effectiveness of the fused feature extraction and ML-KNN approach for multi-label protein localization.

Conclusions:

• The combined feature extraction and ML-KNN approach significantly improves multi-label protein subcellular localization prediction.
• This method offers a more accurate and efficient alternative to traditional experimental techniques.
• Accurate prediction of multi-location proteins has important implications for basic biology and bioinformatics research.