| Title: |
Utilizing Machine Learning Approaches to Understand the Interrelationship of Diet, the Human Gastrointestinal Microbiome, and Health |
| Authors: |
Guetterman, Heather; Auvil, Loretta; Russell, Nate; Welge, Michael; Berry, Matt; Gatzke, Lisa; Bushell, Colleen; Holscher, Hannah |
| Source: |
The FASEB Journal ; volume 30, issue S1 ; ISSN 0892-6638 1530-6860 |
| Publisher Information: |
Wiley |
| Publication Year: |
2016 |
| Collection: |
Wiley Online Library (Open Access Articles via Crossref) |
| Description: |
Background A growing body of literature supports the ability of specific foods and nutrients to impact the gastrointestinal microbiome. However, there is a dearth of knowledge on the interplay of dietary components (e.g. foods and nutrients), gastrointestinal bacteria, and bacterial metabolites. Current analytical approaches limit investigation of these complex interrelations; therefore, further research utilizing modern machine learning methods is needed. Objective We aimed to fill the gap in knowledge about the interrelationship of diet, the gastrointestinal microbiome, and health by utilizing multivariate approaches that address P>>N, many features but few samples to 1) validate results generated by prototype software with previously published results; and 2) identify novel associations among relevant foods, nutrients, bacteria, and bacterial metabolites. Methods Data generated from a human dietary intervention study that included habitual food and nutrient intake patterns (NHANES food frequency questionnaire), daily dietary intake records, breath gas, and fecal bacteria and metabolite data were analyzed using machine learning approaches. Relevant bacteria, bacterial metabolites, foods, and nutrients were identified using Random Forest and relationships among these relevant features were determined using Maximal Information Coefficient. Results Breath hydrogen and agave inulin supplementation were predictive of Bifidobacterium abundance. Habitual diet factors were highly relevant in predicting participant body mass index (BMI). Among bacteria, Phascolarctobacterium , Collinsella , and Erysipelotrichacea were important features for predicting BMI. Oscillospira was associated with habitual dietary glycemic load. Phascolarctobacterium and Eubacterium were associated with habitual intake of deep yellow vegetables. Dietary polyunstaruated fatty acids were relevant in predicting Bacteroidetes to Firmicutes ratios. Total fiber and butyrate were highly relevant in predicting Faecalibacterium abundance. ... |
| Document Type: |
article in journal/newspaper |
| Language: |
English |
| DOI: |
10.1096/fasebj.30.1_supplement.406.3 |
| Availability: |
https://doi.org/10.1096/fasebj.30.1_supplement.406.3 |
| Rights: |
http://onlinelibrary.wiley.com/termsAndConditions#vor |
| Accession Number: |
edsbas.D9DAB429 |
| Database: |
BASE |