iPPBS-PseAAC

iPPBS-PseAAC predicts protein-protein binding sites from amino acid sequences by extracting PseAAC-based features transformed with stationary wavelet transform for ensemble classification.

Key Features:

• 15-Tuple Peptide Segments: Each amino acid residue is represented as a 15-tuple peptide segment via a sliding window centered on the target residue to capture local sequence context.
• Pseudo Amino Acid Composition (PseAAC): The general form of PseAAC converts each peptide segment into a numerical series based on physicochemical properties.
• Stationary Wavelet Transform: The numerical series is transformed by stationary wavelet transform into a 20-dimensional feature vector that highlights low-frequency internal motions.
• Two-layer Ensemble Classifier: A two-layer ensemble system is used where the first layer employs multiple Random Forest classifiers across bootstrap-derived training sets and the second layer selects the most relevant physicochemical property from seven available options.

Scientific Applications:

• Structural biology: Identification of protein-protein binding sites to inform structural studies and interpretations of internal motions.
• Drug design: Prioritization of interface residues relevant to small-molecule or biologic design.
• Molecular docking studies: Selection of candidate binding-site residues for docking and interaction modeling.
• Functional annotation: Sequence-based annotation of residues involved in protein-protein interactions.
• Bioinformatics research: Development and benchmarking of sequence-based PPBS prediction methods.

Methodology:

Residue-centered 15-tuple sliding-window segments are encoded by the general form of PseAAC into numerical series, transformed by stationary wavelet transform into a 20-dimensional feature vector, and classified by a two-layer ensemble (first-layer Random Forests over bootstrap training sets; second-layer selection among seven physicochemical properties).