Allerdictor

Allerdictor predicts protein allergenicity by modeling protein sequences as text and applying support vector machines (SVM) for text classification to identify potential allergens in large sequence databases.

Key Features:

• Speed and scale: Scans large protein databases such as Swiss-Prot (~540,000 sequences) in approximately 6 minutes on a single-core PC.
• High precision and recall: Maintains high precision at elevated recall levels on highly skewed datasets, reducing false positives in allergen detection.
• Text-based sequence modeling: Treats protein sequences as text documents to enable text-classification approaches.
• Machine learning classifier: Utilizes support vector machines (SVM) for classification of protein-derived textual features.
• Alignment-free approach: Circumvents traditional sequence alignment methods, reducing computational intensity.
• Batch processing: Supports large-scale batch processing of sequences.
• Implementation: Implemented in Python.

Scientific Applications:

• Biotechnology and food safety: Identifies potential allergenic proteins for assessment and removal in biotechnology-derived products and food safety pipelines.
• Genomic annotation: Annotates sequenced genomes for potential allergenic proteins to support studies of genetic bases of allergies.
• Public health surveillance: Enhances allergen identification to support public health initiatives aimed at reducing allergen exposure and managing allergic risks.

Methodology:

Models protein sequences as text documents and applies support vector machines (SVM) for text classification in an alignment-free framework; demonstrated by scanning the Swiss-Prot database (~540,000 sequences) in ~6 minutes on a single-core PC.