SLPred

SLPred predicts subcellular localizations of human proteins using an ensemble-based multi-view, multi-label approach to assign nine main subcellular locations and support functional interpretation of protein distribution.

Key Features:

• Ensemble multi-view, multi-label architecture: Uses an ensemble-based multi-view framework to perform multi-label prediction of protein subcellular localizations.
• Per-location models: Employs independent machine-learning models trained specifically to predict each of nine main subcellular locations.
• Curated training data: Leverages curated annotations from UniProtKB/Swiss-Prop human protein entries for training.
• Ontology-aware hierarchy integration: Integrates disjoint terms within the UniProt subcellular localization hierarchy by referencing relationships defined in the cellular component category of Gene Ontology.
• Benchmark evaluation: Performance has been evaluated on multiple benchmarking datasets, including proprietary in-house sets, and compared against six state-of-the-art methods.

Scientific Applications:

• Functional annotation: Assigns subcellular localizations to support functional annotation of proteins.
• Proteomics studies: Supports proteomics analyses that require protein compartmentalization information.
• Disease mechanism investigation: Aids studies linking protein mislocalization to disease mechanisms.
• Drug target discovery: Provides localization data relevant to identifying and prioritizing drug targets.

Methodology:

Ensemble-based multi-view, multi-label prediction using independent machine-learning models trained per subcellular location on curated UniProtKB/Swiss-Prop human protein annotations, with integration of UniProt SL hierarchy terms via Gene Ontology cellular component relationships and evaluation against multiple benchmarking datasets including proprietary in-house sets and six state-of-the-art methods.