SUBAcon

SUBAcon predicts consensus subcellular localizations for Arabidopsis thaliana proteins by integrating multiple computational predictors and experimental datasets.

Key Features:

• Naive Bayes Classifier Framework: SUBAcon employs a naive Bayes classifier to integrate heterogeneous evidence and produce probabilistic outputs.
• Integration of Multiple Data Sources: Integrates 22 computational prediction algorithms together with green fluorescent protein (GFP) tagging data, mass spectrometry (MS) localization data, protein-protein interaction networks, and co-expression data.
• Consensus Call and Probability Derivation: Synthesizes evidence to generate a consensus subcellular localization call per protein with an associated probability score.
• Improved Accuracy Over Single Predictors: Combines diverse evidence to classify protein locations with higher accuracy than individual prediction methods.
• Reference Datasets: Leverages the SUBA3 database and the ASURE training set as part of its evidence base.

Scientific Applications:

• Proteome-wide Localization Studies: Enables mapping of subcellular localizations across the Arabidopsis proteome for studies of protein function and compartmentalization.
• Biological Network Development: Provides localization constraints that support construction and interpretation of cellular biological networks and interaction maps.

Methodology:

SUBAcon uses a naive Bayes classifier to integrate outputs from 22 computational predictors with GFP tagging and MS localization data, protein-protein interaction networks, and co-expression data to produce consensus localization calls with associated probability scores.