Data integration for prediction of weight loss in randomized controlled dietary trials

Ruby I Nielsen ,Lea Benedicte Skov Hansen,Thomas Nordahl Petersen,Hanne Frøkiær,Anders Christensen ,Rikke Juul Gøbel ,Marlene Danner Dalgaard,Derya Aytan-Aktug,Karsten Kristiansen,Vincent Aaskov,Josef Korbinian Vogt ,Susanne Brix,Rasa A Muktupavela ,Oluf Pedersen,Sara Luis García ,Torben Hansen,Henrik Munch Roager ,Lotte Lauritzen,Mette Kristensen,Tine Rask Licht,Henrik Vestergaard,Cecilia Bang Jensen,Mads Vendelbo Lind,Martin Iain Bahl ,Morten H Sparholt,Christina Warinner,Marianne Helenius,Ramneek Gupta

doi:10.1038/s41598-020-76097-z

Abstract

Diet is an important component in weight management strategies, but heterogeneous responses to the same diet make it difficult to foresee individual weight-loss outcomes. Omics-based technologies now allow for analysis of multiple factors for weight loss prediction at the individual level. Here, we classify weight loss responders (N = 106) and non-responders (N = 97) of overweight non-diabetic middle-aged Danes to two earlier reported dietary trials over 8 weeks. Random forest models integrated gut microbiome, host genetics, urine metabolome, measures of physiology and anthropometrics measured prior to any dietary intervention to identify individual predisposing features of weight loss in combination with diet. The most predictive models for weight loss included features of diet, gut bacterial species and urine metabolites (ROC-AUC: 0.84–0.88) compared to a diet-only model (ROC-AUC: 0.62). A model ensemble integrating multi-omics identified 64% of the non-responders with 80% confidence. Such models will be useful to assist in selecting appropriate weight management strategies, as individual predisposition to diet response varies.

Highlights

Diet is an important component in weight management strategies, but heterogeneous responses to the same diet make it difficult to foresee individual weight-loss outcomes
102 participants completed the whole grain (N = 50 participants) and gluten (N = 52 participants) clinical trials. Both trials reported significant weight loss following each of the interventions compared to a refined grain d iet[25,26]
An individual with any degree of weight loss after the 8-week intervention compared to baseline body weight was considered a responder, whereas no change or weight gainers were classified as non-responders independent of the dietary study arm

Summary

Introduction

Diet is an important component in weight management strategies, but heterogeneous responses to the same diet make it difficult to foresee individual weight-loss outcomes. Random forest models integrated gut microbiome, host genetics, urine metabolome, measures of physiology and anthropometrics measured prior to any dietary intervention to identify individual predisposing features of weight loss in combination with diet. A model ensemble integrating multi-omics identified 64% of the nonresponders with 80% confidence Such models will be useful to assist in selecting appropriate weight management strategies, as individual predisposition to diet response varies. We present random forest-based data integration of anthropometry, blood serum markers, gut microbiome markers, urine metabolomics and host genomics to investigate, if the weight loss response can be predicted based on randomisation onto dietary intervention and biomarkers at baseline prior to any dietary intervention. Understanding predictive features of weight loss response will drive improved understanding of the interplay between gut microbiota, diet and individual predisposition

Methods

Results

Conclusion