Signal-3L

Signal-3L predicts signal peptide sequences and their cleavage sites in protein sequences from eukaryotic and bacterial origins to identify secretory proteins for applications such as drug discovery and gene therapy.

Key Features:

• Three-layer prediction architecture: Employs three distinct prediction engines that operate at progressively deeper layers for classification and cleavage-site identification.
• Secretory vs. Non-Secretory Classification: Uses an ensemble classifier constructed by fusing multiple OET-KNN (optimized evidence-theoretic K nearest neighbor) classifiers operating within various dimensions of PseAA (pseudo amino acid) composition spaces.
• Candidate selection for cleavage sites: Applies a subsite-coupled discrimination algorithm to select potential signal peptide cleavage-site candidates in secretory proteins.
• Final cleavage-site determination: Integrates global sequence alignment outcomes for each candidate through a voting system to determine the precise cleavage site.
• Organism scope: Handles protein sequences from both eukaryotic and bacterial origins.
• Swiss-Prot-derived outputs: Produces lists of signal peptides identified from Swiss-Prot entries lacking annotations or containing uncertain terms, compiled as Tab-Signal-3L.xls and updated annually.
• Prediction performance: Exhibits high success prediction rates and efficient computational performance suitable for large-scale dataset analyses.

Scientific Applications:

• Drug discovery: Identification of secretory proteins and their signal peptides to support target discovery and therapeutic development.
• Gene therapy: Characterization of signal peptides to inform design and delivery of therapeutic proteins.
• Large-scale proteome analysis: High-throughput annotation of signal peptides and cleavage sites across extensive protein datasets.
• Proteome annotation: Recovery and annotation of signal peptides in Swiss-Prot entries that lack explicit signal peptide annotations or contain uncertain terms.

Methodology:

Computational steps include (1) classifying proteins as secretory or non-secretory using an ensemble of OET-KNN classifiers in various PseAA composition dimensions, (2) selecting candidate cleavage sites with a subsite-coupled discrimination algorithm, and (3) determining the final cleavage site by integrating global sequence alignment results via a voting system.