RAACBook

RAACBook provides reduced amino acid alphabet-based analysis of protein primary sequences to support sequence analysis and machine learning by simplifying sequences to improve computational efficiency, minimize information redundancy, and reduce overfitting.

Key Features:

• Comprehensive Reduced Amino Acid Alphabet Library: Integrates 74 types of reduced amino acid alphabets manually curated into 673 distinct reduced amino acid clusters (RAACs).
• Sequence analysis with K-tuple, g-gap, and λ-correlation: Computes K-tuple reduced amino acid compositions from protein primary sequences using K-tuple, g-gap, and λ-correlation parameters.
• Visualization outputs: Produces sequence alignment, mergence of RAA composition, feature distribution, and logos of the reduced sequence.
• Machine learning-based protein classification: Trains classification models using K-tuple RAACs and reports performance metrics including ROC, AUC, and MCC.

Scientific Applications:

• Sequence-dependent inference and 3D modeling: Supports sequence-dependent inference relevant to 3D protein structure modeling as exemplified by AlphaFold.
• Protein classification and predictive modeling: Provides RAAC-based feature generation for protein classification and predictive modeling to investigate structure–function relationships.

Methodology:

Uses 74 reduced amino acid alphabets forming 673 RAACs; computes K-tuple reduced amino acid compositions with K-tuple, g-gap, and λ-correlation parameters; generates sequence alignment, merged RAA composition, feature distribution, and reduced-sequence logos; trains classifiers evaluated by ROC, AUC, and MCC.