HSLpred

HSLpred predicts the subcellular localization of human proteins, assigning them to cytoplasmic, mitochondrial, nuclear, or plasma membrane compartments to support functional and interaction inference.

Key Features:

• Support Vector Machine (SVM) classifiers: Uses Support Vector Machine classifiers to perform localization prediction.
• Amino Acid Composition: Uses amino acid composition, achieving 76.6% accuracy.
• Dipeptide Composition (i + 1): Uses dipeptide composition (i + 1) with 77.8% accuracy.
• Higher-order Dipeptides (i + 2, i + 3, i + 4): Constructs SVM modules using higher-order dipeptides (i + 2, i + 3, i + 4) with accuracies of 79.7%, 77.5%, and 77.1%, respectively.
• PSI-BLAST Similarity Search: Employs PSI-BLAST against a nonredundant database of experimentally annotated proteins, yielding 73.3% accuracy.
• Hybrid Modules: Implements hybrid modules: Hybrid1 combines amino acid composition, dipeptide (i + 1) and PSI-BLAST (84.9%); Hybrid2 combines dipeptide (i + 1) and higher-order dipeptides (i + 2, i + 3, i + 4) (81.3%); Hybrid3 combines amino acid composition, dipeptide (i + 1), higher-order dipeptides and PSI-BLAST (84.4%).

Scientific Applications:

• Protein function annotation: Supports inference of protein function by assigning proteins to cytoplasmic, mitochondrial, nuclear, or plasma membrane compartments.
• Interaction and pathway analysis: Aids study of protein interactions and pathways within specific cellular compartments.
• Proteome annotation: Assists annotation of newly discovered human proteins in proteomics studies.
• Disease mechanism investigation: Facilitates investigation of disease mechanisms related to protein mislocalization.

Methodology:

Applies Support Vector Machine classifiers trained on amino acid composition, dipeptide (i + 1) and higher-order dipeptide (i + 2, i + 3, i + 4) features and integrates PSI-BLAST similarity searches against a nonredundant database of experimentally annotated proteins, with hybrid SVM modules combining these features.