Allerdictor: fast allergen prediction using text classification techniques.

Area of Science:

• Biotechnology
• Bioinformatics
• Genomics

Background:

• Accurate identification and removal of allergens from biotechnology products are crucial for human health.
• Existing allergen prediction tools often produce false positives in large-scale genomic analyses due to data imbalance.
• Current methods are slow and lack efficient batch processing capabilities, hindering effective allergen discovery.

Purpose of the Study:

• To develop a fast and accurate sequence-based allergen prediction tool for large-scale genomic analysis.
• To address the limitations of existing allergen prediction tools, including low precision and inefficiency.

Main Methods:

• Developed Allerdictor, a tool that treats protein sequences as text documents.
• Employed support vector machine (SVM) in text classification for allergen prediction.
• Utilized a standalone and web server implementation for accessibility and batch submission.

Main Results:

• Allerdictor demonstrated high precision and high recall on skewed datasets.
• The tool achieved fast prediction speeds, scanning the Swiss-Prot database in approximately 6 minutes on a single PC.
• Identified less than 1% of scanned sequences as potential allergens.

Conclusions:

• Allerdictor provides an effective and efficient solution for large-scale allergen discovery.
• The tool's speed and accuracy improve upon existing methods for identifying potential allergens in genomic data.
• Allerdictor enhances the capability for public domain allergen discovery.