AtSubP

AtSubP predicts subcellular localization of Arabidopsis thaliana proteins to support proteome annotation and functional interpretation.

Key Features:

• Integrative approach: Employs a support vector machine-based methodology that integrates amino acid composition, sequence-order effects, terminal information, Position-Specific Scoring Matrix (PSSM), and similarity search data from the Position-Specific Iterated-Basic Local Alignment Search Tool (PSI-BLAST).
• High accuracy: Achieved 91% overall sensitivity in 5-fold cross-validation for seven subcellular compartments with precision of 90.9% and Matthews correlation coefficient of 0.89.
• Species-specific optimization and benchmarking: Benchmarking against independent Swiss-Prot datasets and proteins identified by green fluorescent protein tagging and mass spectrometry shows performance superior to TargetP, LOCtree, PA-SUB, MultiLoc, WoLF PSORT, Plant-PLoc, and All-Plant, while performance is reduced on non-trained organisms such as rice, soybean, human, yeast, fruit fly, and worm.

Scientific Applications:

• Proteome annotation: Provides precise subcellular localization predictions to aid comprehensive annotation of the Arabidopsis proteome.
• Functional genomics: Links localization predictions to potential biological roles and regulatory mechanisms for functional genomics studies.
• Comparative analyses: Enables benchmarking and comparison of localization prediction methods across datasets and species, highlighting Arabidopsis-specific model performance.

Methodology:

Support vector machine-based hybrid classifier integrating amino acid composition, sequence-order effects, terminal information, PSSM and PSI-BLAST similarity search; evaluated by 5-fold cross-validation and benchmarked against independent Swiss-Prot and experimentally identified (GFP tagging, mass spectrometry) protein datasets.