AllergenFP

AllergenFP predicts allergenicity of protein sequences using an alignment-free descriptor fingerprint approach to classify allergens and non-allergens.

Key Features:

• Alignment-Free Approach: AllergenFP employs a descriptor-based alignment-free methodology that does not rely on sequence alignment for allergen identification.
• Four-Step Algorithm: Four-step algorithm: descriptor generation from amino acid principal properties (hydrophobicity, size, relative abundance, helix forming propensity, β-strand forming propensity), auto- and cross-covariance (ACC) transformation to equal-length vectors, conversion to binary fingerprints, and similarity assessment via the Tanimoto coefficient.
• Accuracy and Benchmarking: Benchmarked on a dataset of 2,427 allergens and 2,427 non-allergens, it achieved 88% accuracy and a Matthews correlation coefficient of 0.759.
• E-descriptors Property Capture: The set of E-descriptors captures main structural and physicochemical properties of amino acids to enhance predictive capability.
• Universality: The descriptor fingerprint approach is applicable to various classification problems in computational biology beyond allergenicity prediction.

Scientific Applications:

• Allergenicity Prediction: Classification of protein sequences for their potential to cause allergic reactions.
• Protein Classification: General protein classification tasks and other classification problems in computational biology using descriptor-based fingerprints.

Methodology:

Descriptor generation using E-descriptors based on amino acid properties, auto- and cross-covariance (ACC) transformation to produce equal-length vectors, conversion to binary fingerprints, and comparison using the Tanimoto coefficient.