Making sense of big data can empower us to ‘look’ into the future. Take the weather app on your phone, for instance; it forecasts rainy days based on atmospheric patterns, historical weather trends and real-time meteorological data.
Similarly, scientists have been working on a ‘weather app’ to forecast protein allergens: substances in everyday foods and other consumables that might, in an unlucky minority of people, cause immune system flare-ups. Also known as allergic reactions, these flare-ups can range from mild but irritating rashes to life-threatening breathing problems.
Whether you’re a food producer, regulator or consumer, having access to tools to predict a protein’s allergenicity—how likely it is to trigger a bad immune response—would be hugely beneficial. Existing computational platforms for mapping and predicting allergenicity are, however, notoriously unreliable, with estimated accuracy levels between 51 to 73 percent at best.
According to Sebastian Maurer-Stroh, Deputy Executive Director at A*STAR’s Bioinformatics Institute (BII), data is the key to bumping up the dependability of these predictive platforms. With that in mind, Maurer-Stroh and colleagues built AllerCatPro 2.0, a computational platform that may set a new gold standard for allergenicity prediction.
“AllerCatPro 2.0 predicts a protein’s allergenic potential based on the most comprehensive international databases of proteins reliably associated with allergenicity,” said Maurer-Stroh, listing examples such as the WHO/International Union of Immunological Societies (WHO/IUIS) and the Comprehensive Protein Allergen Resource (COMPARE).
What sets the platform apart from its predecessors is its ability to crunch data from large, multidimensional datasets that include amino acid sequences and 3D protein structures, thereby boosting its accuracy when pinpointing potentially problematic proteins.
In their study, the international team of researchers deployed AllerCatPro 2.0 to compile a list of nearly 5,000 known protein allergens, including 165 human proteins with links to autoimmune diseases. They also identified 162 safer protein options with low allergenic scores. When tested, AllerCatPro 2.0 significantly outperformed other benchmark platforms in terms of reliability, scoring an impressive 84 percent accuracy.
“AllerCatPro 2.0 can be used as the first step in food safety assessments to find potential new allergens that might warrant further evaluation,” explained Maurer-Stroh, adding that the team is working with industry collaborators to continually update the platform as new allergens are identified.
Usability was also a priority in designing the platform, which is currently freely accessible as a web server. “We created a user-friendly interface with updated features that provides detailed results on potential cross-reactivity, protein information, functionality, clinical relevance and information relating to similar allergens,” Maurer-Stroh said.
The team is currently expanding the platform’s capabilities for detecting allergens in specific fish species; and is also working together with academic and industrial collaborators on a project to assess food safety risks in Singapore’s urban aquaculture industry.
The A*STAR-affiliated researchers contributing to this research are from the Bioinformatics Institute (BII) and the Singapore Institute for Food and Biotechnology Innovation (SIFBI).