PSORT.org

PSORT.org predicts subcellular localization of proteins using the PSORT family algorithms (including PSORTb and WoLF PSORT) to inform molecular biology and genome annotation.

Key Features:

• PSORTb for Gram-negative bacteria: Predicts localization across five sites—cytoplasm, inner membrane, periplasm, outer membrane, and extracellular space—and returns associated probability scores.
• Sequence attributes analyzed: Evaluates amino acid composition, similarity to proteins with known localizations, presence of signal peptides, transmembrane alpha-helices, and specific motifs.
• Probabilistic integration: Integrates multiple sequence analyses via a probabilistic method to generate localization predictions with confidence scores.
• Rule-based expert system: Uses expert "if-then" rules derived from experimental observations to identify hydrophobic stretches, signal sequences, and consensus patterns for lipoproteins.
• Performance metrics: Reports 97% precision and 75% recall based on 5-fold cross-validation on a dataset of 1,443 experimentally localized proteins.
• Comparative classifier analysis: Methodologies are informed by comparative studies involving k nearest neighbors, decision trees, and naïve Bayes classifiers with demonstrated efficacy for organisms such as yeast and E. coli.
• PSORT family inclusion: Provides implementations from the PSORT family, explicitly including PSORTb and WoLF PSORT.

Scientific Applications:

• Genome annotation: Infers protein function and cellular roles by predicting subcellular localization to support annotation of genomes.
• Drug discovery: Identifies proteins localized to specific subcellular compartments as potential targets for drug development against Gram-negative bacteria.

Methodology:

Combines rule-based "if-then" expert systems and probabilistic integration of sequence analyses, with performance assessed by 5-fold cross-validation and comparative evaluations against k nearest neighbors, decision trees, and naïve Bayes classifiers.