Novel machine learning method allerStat identifies statistically significant allergen-specific patterns in protein...

Area of Science:

• Food science and technology
• Bioinformatics
• Molecular biology

Background:

• Genome editing and synthetic biology enable novel food and protein production.
• Accurate evaluation of food and protein toxicity and allergenicity is crucial.
• Known allergen-specific patterns (ASPs) are limited, necessitating new discovery methods.

Purpose of the Study:

• To introduce a data-driven, machine learning approach for discovering novel allergen-specific patterns (ASPs) in amino acid sequences.
• To enable an exhaustive search for amino acid subsequences with significantly higher frequencies in allergenic proteins.
• To improve the prediction of protein allergenicity for novel food and protein safety assessments.

Main Methods:

• Developed a data-driven approach utilizing machine learning to identify allergen-specific patterns (ASPs).
• Conducted an exhaustive search for amino acid subsequences overrepresented in allergenic proteins.
• Created and applied the method to a database of 21,154 proteins with known allergenicity data.

Main Results:

• Identified novel allergen-specific patterns (ASPs) consistent with existing biological knowledge.
• Achieved higher allergenicity prediction performance compared to existing methods.
• Demonstrated the utility of the detected ASPs in evaluating synthetic food and protein safety.

Conclusions:

• The proposed data-driven machine learning method effectively discovers novel allergen-specific patterns (ASPs).
• This approach enhances the accuracy of allergenicity prediction for proteins.
• The method shows promise for evaluating the safety of novel synthetic foods and functional proteins.