PSLpred: Prediction of Subcellular Localization of Bacterial Proteins

PSLpred: Prediction of Subcellular Localization of Bacterial Proteins

PSLpred is a computational web server developed for predicting the subcellular localization of Gram-negative bacterial proteins.

The tool predicts whether a bacterial protein belongs to cytoplasmic, extracellular, inner-membrane, outer-membrane, or periplasmic localization classes. PSLpred uses a hybrid machine learning approach that combines PSI-BLAST similarity search with support vector machine models based on amino acid composition, dipeptide composition, and physicochemical properties.

Web Server: http://www.imtech.res.in/raghava/pslpred/

Citation

Bhasin, M., Garg, A., and Raghava, G. P. S. PSLpred: prediction of subcellular localization of bacterial proteins. Bioinformatics, 21(10), 2522-2524, 2005.

https://doi.org/10.1093/bioinformatics/bti309

About the Research

Subcellular localization prediction is important because the biological function of a protein is strongly linked to where it is located inside or outside the cell.

In Gram-negative bacteria, proteins may localize to different compartments such as cytoplasm, inner membrane, periplasm, outer membrane, or extracellular space. Correct prediction of these locations helps in genome annotation, functional annotation, virulence factor identification, and drug or vaccine target discovery.

Earlier bacterial localization tools such as PSORT I, PSORT-B, NNPSL, and CELLO showed useful performance, but there was still scope for improvement, especially for extracellular proteins. PSLpred was developed to improve the prediction accuracy for Gram-negative bacterial protein localization using a hybrid SVM-based strategy.

Data Compilation: The dataset used in this study was the same dataset used previously for CELLO and PSORT-B. It was generated from SWISS-PROT release 40.29 and originally contained 1443 proteins. After removing 141 proteins with more than one localization, 1302 proteins were used for model development.

The final dataset contained:

248 cytoplasmic proteins

268 inner-membrane proteins

244 periplasmic proteins

352 outer-membrane proteins

190 extracellular proteins

Methodology: PSLpred uses support vector machine models trained with amino acid composition, dipeptide composition, physicochemical properties, and PSI-BLAST output. A hybrid model combining all these features achieved the best performance.