SecretomeP

SecretomeP predicts non-classical protein secretion in mammalian and bacterial proteomes by identifying proteins lacking N-terminal signal peptides that are secreted via alternative extracellular routes.

Key Features:

• Sequence-Based Prediction: Employs sequence-based analysis to identify mammalian proteins lacking N-terminal signal peptides, including extracellular matrix proteins such as fibroblast growth factors, interleukins, and galectins.
• Pathway-Independent Features: Detects shared features among secreted proteins independent of classical secretion pathways, including cases where only the mature part is annotated or signal peptides remain uncleaved.
• Comprehensive Proteome Scanning: Scans entire proteomes to identify candidate human proteins potentially undergoing non-classical secretion.
• Bacterial Non-Classical Secretion Prediction: Compiles extracellular bacterial proteins lacking signal peptides, identifies potential "moonlighting" proteins with dual functions, and uses pattern recognition to explore putative motifs correlated with secretion.
• Structural Analysis: Analyzes amino acid composition, secondary structure, and disordered regions to distinguish non-classically secreted bacterial proteins from cellular proteins, noting higher structural disorder in secreted proteins.
• Artificial Neural Networks: Uses artificial neural networks to build feature-based predictive models for non-classical secretion in Gram-positive and Gram-negative bacteria.

Scientific Applications:

• Identification of Novel Proteins: Predicts candidate non-classically secreted proteins to support discovery of novel extracellular proteins in mammals and bacteria.
• Understanding Protein Functionality: Elucidates multifunctionality and "moonlighting" behavior of proteins that have both cytoplasmic and extracellular roles.
• Proteomic Research: Supports proteomic studies of secretion mechanisms beyond signal peptide-dependent pathways.

Methodology:

Analyzing amino acid sequences to identify shared features among known non-classically secreted proteins; evaluating amino acid composition, secondary structure, and disorder; applying artificial neural networks to construct predictive models; and scanning entire proteomes to uncover candidate non-classically secreted proteins.