Integrating Metagenomic Information into Personalized Nutrition Tools: The PREDICT I Study (P20-005-19). (13th June 2019)

• Record Type: Journal Article
• Authors: Spector, Tim; Berry, Sarah; Valdes, Ana; Drew, David; Chan, Andrew; Franks, Paul; Asnicar, Francesco; Segata, Nicola; Davies, Richard
• Is Part Of: Current developments in nutrition. Volume 3(2019)Supplement 1
• Publication Date: 2019-06-13
• Subjects: Nutrition -- Periodicals; Nutritional Physiological Phenomena; Nutrition; Periodicals; Periodicals; Fulltext; Internet Resources; Periodicals; 612.3
• DOI: 10.1093/cdn/nzz040.P20-005-19
• Languages: English
• ISSNs: 2475-2991

Abstract

Abstract: Objectives: The existence of a link between the intestinal microbiome and diet is well established. The demonstration that the microbiome information increases the prediction accuracy of postprandial blood glucose levels (Zeevi et al, 2015) is opening intriguing perspectives for developing personalized nutrition tools. However, reproducibly inferring the diet-induced microbiome changes and stratifying individual responses to dietary interventions based on the microbiome remain open challenges. The PREDICT I study aims to develop: (i) a protocol for gut microbiome sampling and analysis for large-scale nutritional studies and (ii) a microbiome-based machine learning integrative component for predictive personalized nutrition tools. Methods: We performed three metagenomic investigations to; (i) identify the best combination for stool collection, sample storage, DNA extraction, and sequencing ( n = 45); (ii) develop and validate the computational pipeline on an exploratory dietary interventional cohort ( n = 1000); (iii) apply the validated pipeline on an independent validation cohort ( n = 100). The generated total dataset (>8x10^12 sequenced bases) was analyzed with existing and newly developed computational tools and integrated with the metagenomic profiles of >10, 000 samples processed from public repositories. Results: Our resulting validated protocol involves a minimally time-demanding procedure for at-home sample collection, sample storage in a preservation buffer, and DNA extraction with a recently commercialized kit (Qiagen). Metagenomic sequencing proved substantially more accurate than 16S rRNA sequencing and was able to perfectly capture subject-specific strain-level features with longitudinal sampling. This method was also able to stratify by pre-intervention habitual dietary regimes. Our prediction algorithm showed that embedding the microbiome features in a 50-dimension space was sufficient to improve the prediction performance of postprandial blood glucose levels. Conclusions: We present the largest investigation to date on the reproducible connections between the gut microbiome and dietary interventions. Further we describe our methods and results in using the microbiome as a component of a precise integrated postprandial blood glucose and blood lipid level predictor. Funding Sources: Zoe Global Limited, National Institute for Health Research (NIHR), Wellcome Trust.