PSLpred

PSLpred predicts the subcellular localization of proteins from Gram-negative bacterial genomes to support studies of protein function and cellular compartmentalization.

Key Features:

• Hybrid methodology: Integrates PSI-BLAST with three Support Vector Machine (SVM) modules.
• SVM feature sets: SVM modules use residue composition, dipeptide sequences, and physico-chemical property features.
• Overall accuracy: Achieves 91.2% prediction accuracy across Gram-negative bacterial proteins.
• Class-wise accuracy: Cytoplasmic 90.7%, Extracellular 86.8%, Inner-membrane 90.3%, Outer-membrane 95.2%, Periplasmic 90.6%.
• Reliability index performance: Predicts approximately 74% of sequences with an average accuracy of 98% at a recall index (RI) of 5.

Scientific Applications:

• Microbial genomics: Aids annotation of protein subcellular localization in Gram-negative bacterial genomes.
• Functional inference: Supports inference of protein function and interactions based on compartmentalization.
• Physiology and pathogenesis studies: Informs studies of bacterial physiology and mechanisms of pathogenesis.
• Drug and vaccine target identification: Assists in prioritizing proteins for drug discovery and vaccine development.
• Microbial ecology: Supports analyses of protein roles in environmental and ecological contexts.

Methodology:

Integrates PSI-BLAST with three SVM modules trained on residue composition, dipeptide composition, and physico-chemical property features.