SubCellProt: predicting protein subcellular localization using machine learning approaches.

Area of Science:

• Bioinformatics
• Computational Biology
• Proteomics

Background:

• High-throughput genome sequencing generates vast amounts of unannotated raw sequence data.
• Protein function is often deciphered by determining its subcellular localization.
• Experimental methods for proteome annotation are costly and labor-intensive.

Purpose of the Study:

• To present in silico methods as an alternative for predicting protein subcellular localization.
• To develop and evaluate two machine learning approaches for predicting subcellular localization from primary sequence information.
• To enable high-throughput proteome annotation.

Main Methods:

• Utilized two machine learning algorithms: k-Nearest Neighbor (k-NN) and Probabilistic Neural Network (PNN).
• Classified unknown proteins into one of 11 subcellular localizations based on primary sequence information.
• Employed a consensus approach combining predictions from both algorithms and assigned a probability score.

Main Results:

• Primary sequence-derived features, including amino acid composition, sequence order, and physicochemical properties, can predict subcellular localization with fair accuracy.
• The developed machine learning models demonstrate enhanced accuracy in predicting protein subcellular localization.
• The method is suitable for high-throughput proteome annotation.

Conclusions:

• In silico prediction of protein subcellular localization using primary sequence information is feasible and accurate.
• The combination of k-NN and PNN algorithms provides a robust approach for proteome annotation.
• The SubCellProt tool facilitates high-throughput analysis of protein localization.