LB015 - A NOVEL METHODOLOGY TO IDENTIFY FOOD COMPOUNDS ASSOCIATED WITH THE DEVELOPMENT OF INFLAMMATORY BOWEL DISEASE

LB015

A NOVEL METHODOLOGY TO IDENTIFY FOOD COMPOUNDS ASSOCIATED WITH THE DEVELOPMENT OF INFLAMMATORY BOWEL DISEASE

M. Meima^1,2,*, S. Bijlsma¹, M. Meijerink¹, X. Pinho¹, M. Vos¹, F. van Schaik², B. Oldenburg², G. Houben^1,3

¹TNO, ²Gastroenterology and Hepatology, ³Center for Translational Immunology, UMCU, Utrecht, Netherlands

Rationale: We previously developed a method to translate Dutch Food Composition Database (NEVO)-coded dietary data to the level of intakes of food compounds. We now extended our method to European food classification standard FoodEx2-coded dietary data.

Methods: We received dietary data from the European Prospective Investigation into Cancer and Nutrition (EPIC) cohort, including 636 individuals who developed inflammatory bowel disease (IBD) over ~20 years, and 395,273 control subjects. We prepared the dataset for compound-level analysis, which involved cleaning FooDB, the largest online database of food compounds. This cleaning process included removing of incomplete data and inedible parts of food sources, unifying synonyms of food compounds, converting units and recalculating vitamin levels. Hereafter we coupled the FoodEx2-coded dietary data to FooDB according to a previously developed food item matching strategy allowing the calculation of compound intake levels.

Results: Cleaning of FooDB led to the removal of 8,423 compounds due to incomplete data – either lacking a known food source or quantitative values – and an additional 605 compounds were excluded because they occurred only in inedible source components. Using a priority-based matching strategy, we linked 2496 EPIC food items to FooDB to enable calculating compound intake levels (in mg) for ~1500 compounds consumed by EPIC participants. These compounds included common nutrients such as vitamins and minerals, as well as less commonly tracked compounds like polyphenols, terpenes, and atypical fatty acids.

Conclusion: We successfully developed a workflow to calculate food compound intake levels for FoodEx2-coded dietary data from the EPIC-cohort, enabling analyses to identify compounds associated with health and disease. We are currently analyzing these data using both classical statistical approaches and machine learning techniques to investigate associations between specific compounds and the risk of developing IBD.

Disclosure of Interest: M. Meima: None declared, S. Bijlsma: None declared, M. Meijerink: None declared, X. Pinho: None declared, M. Vos: None declared, F. van Schaik Grant / Research Support from: Takeda, Consultant for: Takeda, Galapagos, Speaker Bureau for: Speakers honoraria from Galapagos and Lilly and Janssen-Cilag B.V., Other: Hospitality fees from Ferring and dr. Falk Farma , B. Oldenburg Grant / Research Support from: Galapagos, Abbvie, Takeda, Ferring, Consultant for: Galapagos, Takeda, Janssen, BMS, Pfizer, Abbvie, Ferring, G. Houben: None declared