Pfeature

Pfeature computes over 200,000 sequence- and structure-derived features from protein and peptide amino acid sequences to support residue- and protein-level annotation and predictive modeling.

Key Features:

• Composition-Based Features: Computes traditional and novel compositional descriptors quantifying the elemental makeup of proteins and peptides.
• Binary Profile of Sequences: Calculates fraction and positional information for each amino acid type within a sequence as binary profiles.
• Evolutionary Information-Based Features: Generates position-specific scoring matrix profiles using Position-Specific Iterative Basic Local Alignment Search Tool (PSI-BLAST).
• Structural Descriptors: Computes structural descriptors from protein tertiary structures, including descriptors applicable to non-natural or chemically modified residues.
• Pattern-Based Descriptors: Generates overlapping sequence patterns and computes pattern-based descriptors.
• Model Building: Implements machine learning techniques for classification, regression, and feature selection.

Scientific Applications:

• Structure and Function Annotation: Annotates protein structure and function using computed sequence, evolutionary, and structural features.
• Residue-level Annotation: Supports residue-level annotation via positional, binary, and position-specific scoring matrix profiles.
• Therapeutic Potential Assessment: Assesses therapeutic properties of proteins and peptides using structural descriptors and computed features.
• Analysis of Chemically Modified Peptides: Analyzes chemically modified peptides and residues using computed and structural descriptors.
• Predictive Model Development: Enables development of classification and regression models and feature selection for predictive tasks.

Methodology:

Computational methods explicitly include calculation of compositional descriptors, binary profiles (fraction and positional amino-acid information), generation of overlapping sequence patterns, PSI-BLAST-based generation of position-specific scoring matrix profiles, computation of structural descriptors from tertiary structures, and application of machine learning for classification, regression, and feature selection.