ProFET

ProFET extracts engineered biophysical and sequence-derived features from protein sequences to enable machine learning classification of high-level protein functions without relying on external databases or sequence alignment.

Key Features:

• Comprehensive Feature Extraction: Extracts hundreds of features encompassing biophysical and sequence-derived attributes that capture statistically informative patterns for protein classification.
• Diverse Representations: Generates compact feature sets using various sequence and amino-acid alphabet representations to improve machine learning efficiency.
• Alignment-free Classification: Performs classification without relying on external databases or sequence alignment.
• Universal Application: Applies across datasets to predict subcellular localization, structural classes, and functional properties such as neuropeptide precursors, thermophilic proteins, and nucleic acid-binding proteins.
• Benchmark Performance: Demonstrated robust performance on 17 benchmark datasets for binary and multi-class classification tasks.
• Biological Interpretability: Produces features that provide insight into properties associated with high-level protein functions.
• Scalable Implementation: Implemented in Python and adaptable for multi-genome scale analysis.

Scientific Applications:

• Protein functional annotation: Enables high-level classification of protein functions without database homology or sequence alignment.
• Subcellular localization and structural class prediction: Supports prediction of subcellular localization and structural classes from sequence-derived features.
• Specialized protein identification: Identifies neuropeptide precursors, thermophilic proteins, and nucleic acid-binding proteins from sequence features.
• Drug discovery and functional genomics: Provides feature-based classifications useful for target characterization in drug discovery and analyses in functional genomics and systems biology.

Methodology:

Implements a universal feature-engineering approach that extracts biophysical and sequence-derived attributes using various sequence and amino-acid alphabet representations for machine learning-based classification; implemented in Python.