The Importance of Multi Feature Extraction and Fusion for Prediction of Protein Subcellular Localization

With the advancement of proteomics, the prediction of protein subcellular localization has emerged as a pivotal area of bioinformatics research. The essence of this research lies in the extraction of features from protein sequences and the selection of appropriate prediction algorithms. This article uses three feature extraction methods: pseudo-amino acid composition model, amino acid physicochemical property composition model, and position-specific scoring matrix. The three features and fused features are separately subjected to random forest algorithm. From the prediction accuracy, it can be seen that the importance of multi-feature fusion for protein subcellular feature extraction.