Diets to optimise individual and planetary health (DIPH) – A Machine Learning Study

Food systems (including production, processing, distribution and consumption of foods) are a major driver of greenhouse gas emissions, biodiversity loss and other environmental indicators. They affect all planetary boundaries. At the same time, unhealthy diets are a main driver of diseases. New dietary guidelines such as the EAT-Lancet guidelines or national recommendations from the Nordic countries, Germany or Austria have been established to tackle negative consequences of unhealthy diets for individual and planetary health. However, while these guidelines are based on meta-evidence, there is a lack of studies using individual-level data on actual food consumption and health outcomes to identify sustainable diets that are feasible under real-world conditions.

In our project, we will use data from large-scale population-based studies and machine learning (ML) tools to predict the best possible dietary patterns in terms of co-benefits for individual and planetary health. Our analyses will be based on a comprehensive set of indicators from both health domains to facilitate an integrated prediction of multiple, optimized outcomes by real-world compatible dietary patterns.

You will be part of a multidisciplinary team including epidemiologists, nutrition scientists, sustainability researchers, and computer scientists, with your main affiliation and workplace at Queen’s University Belfast and the Co-Centre for Sustainable Food Systems. The project will be supervised by Prof. Tilman Kühn (Queen’s University Belfast / Co-Centre for Sustainable Food Systems), together with co-supervisors Prof Raymond Bond (Ulster University), Dr Amy Jennings and Prof Aedín Cassidy (Queen’s University Belfast / Co-Centre for Sustainable Food Systems) and Prof Georg Zeller (University of Leiden, NL). You will have the opportunity to attend comprehensive training in advanced ML methods, both locally and internationally. Our goal is to then apply ML to analyse high-dimensional data from large-scale population-based studies. Together with our team, you will help in establishing a database of planetary health indicators. You will use ML to predict the best possible combinations of environmental and health indicators by ideal dietary patterns, using techniques like XGBoost, regression trees, or Ensemble.

Click here to access the application form