SUBA3

SUBA3 provides integrated experimental and computational subcellular localization data for Arabidopsis proteins to assign probabilistic cellular locations.

Key Features:

• Extensive Data Integration: Combines manual literature curation with large-scale subcellular proteomics (mass spectrometry, MS), green fluorescent protein (GFP) visualization, protein-protein interaction (PPI) datasets, co-expression data, and predictions from 22 localization algorithms.
• Enhanced Database Content: Contains over 650,000 new subcellular location calls covering 35,388 non-redundant Arabidopsis proteins.
• SUBAcon Algorithm: Implements a Bayesian approach using a naive Bayes classifier to integrate experimental localizations and computational predictions, including GFP and MS localizations, PPIs, and co-expression data.
• Probabilistic Localization: Provides per-protein probabilities for subcellular compartments to quantify confidence in localization assignments.
• Visualization Tools: Associates localization data with reference organelle datasets, GFP images, and pictographic heat maps representing localization probabilities.

Scientific Applications:

• Protein function inference: Use consensus subcellular localization to inform hypotheses about protein function in Arabidopsis.
• Biological network construction: Support construction and refinement of compartmentalized biological process and interaction networks.
• Evidence reconciliation: Resolve discrepancies among single-method localization results by providing consensus probabilistic calls to improve reliability of Arabidopsis proteomic analyses.

Methodology:

SUBAcon applies a Bayesian naive Bayes classifier to integrate experimental localizations (GFP and MS), PPIs, co-expression data, literature curation, and predictions from 22 algorithms to compute probabilistic consensus subcellular localization calls.