Allergy prediction & prevention

Allergic diseases, such as asthma, are increasing in prevalence, but can be hard to diagnose – especially in young children. The growing availability of multi-omics data (e.g. DNA) opens up exciting new opportunities to better understand, diagnose, and treat these diseases. But unlocking the value of these immense and detailed datasets is a major analytical challenge, due to the complex interplay between genetic and environmental factors.

We integrated multi-omics data to assess the ability of a combination of genetic, epigenetic, and environmental factors to predict allergic disease. To achieve this, we evaluated more than ten AI algorithms, applied various dimensionality reduction methods, and implemented multiple model optimization and evaluation best-practices. To strike the optimal balance between simplicity and efficacy, we developed a predictive model using only 3 epigenetic variables (i.e. DNA methylation sites) that was still able to accurately predict allergic disease in adolescents. The model was successfully replicated with comparable results in an independent cohort of children in Puerto Rico, proving the validity of the model and indicating good generalizability of predicted outcomes.

The research team published in the academic journal Nature Communications in December 2022, receiving 13 citations in the first twelve months, and an impact factor of 16.6 (2023). The findings were also presented at the 2021 conference of the European Respiratory Society. Multiple spin-off research projects have already been initiated, including a program to extend the study to allergic disease in younger children and a precision medicine project that received a Longfonds grant.