Fecal bacteria as biomarkers for predicting food intake in healthy adults

Authors: SHINN, LEILA - University Of Illinois; LI, YUTONG - University Of Illinois; MANSHARAMANI, ADITYA - University Of Illinois; ZHU, RUOQING - University Of Illinois; AUVIL, LORETTA - University Of Illinois; WELGE, MICHAEL - University Of Illinois; BUSHELL, COLLEEN - University Of Illinois; KHAN, NAIMAN - University Of Illinois; Charron, Craig; Novotny, Janet; Baer, David; HOLSCHER, HANNAH - University Of Illinois

Submitted to: Journal of Nutrition
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 8/25/2020
Publication Date: 10/6/2020
Citation: Shinn, L.M., Li, Y., Mansharamani, A., Zhu, R., Auvil, L., Welge, M., Bushell, C., Khan, N.A., Charron, C.S., Novotny Dura, J., Baer, D.J., Holscher, H.D. 2020. Fecal bacteria as biomarkers for predicting food intake in healthy adults. Journal of Nutrition. https://doi.org/10.1093/jn/nxaa285.
DOI: https://doi.org/10.1093/jn/nxaa285

Interpretive Summary: In human nutrition, it is important to develop methods to determine what foods people eat. Typical methods include analyses of blood and urine. To complement these approaches and improve the accuracy in determining food intake, a novel approach using human gastrointestinal microbiota was developed. Microbiota such as bacteria interact in the gastrointestinal tract with the foods that are consumed. To better understand these interactions, a computationally intensive, machine learning approach must be utilized. We aimed to identify bacterial biomarkers that could accurately predict food consumption. Data were combined from five feeding studies that provided diets containing avocados, almonds, broccoli, walnuts, whole grain oats or whole grain barley. Fecal samples were collected at the beginning and end of each diet period and were analyzed by 16S rRNA sequencing to characterize the microbiota. Using the 10 most significant bacterial species from each diet group, we were able to correctly classify the food consumed up to 87% of the time. These approaches and subsequent determination of gastrointestinal bacterial biomarkers may be useful fidelity measures to ascertain compliance in human nutrition studies. Long term, these approaches may lead to diet recommendations to improve health outcomes.

Technical Abstract: Dietary intake affects the human gastrointestinal microbiota. To better understand the host-microbe interactions related to food intake, a more computationally intensive, multivariate, machine learning approach must be utilized. We aimed to identify bacterial biomarkers with high predictive accuracy for dietary intake. Data were aggregated from five randomized, controlled, feeding studies in adults (n=199) that provided diets containing avocados, almonds, broccoli, walnuts, whole grain oats or whole grain barley. Fecal samples were collected during treatment and control periods of each study for 16S rRNA gene (V4 region) sequencing. Sequence data were analyzed using DADA2 and QIIME2. Marginal screening using the Kruskal-Wallis test was performed on all species-level taxa to examine the differences between each of the 6 treatment groups and their respective control groups. The 20 most significant species from each treatment were selected and pooled together for predicting the treatment, i.e., food consumed, using a random forest model. The number of bacterial species was further decreased in a stepwise fashion based on variable importance scores to determine the most compact feature set minimizing the loss in predictive accuracy. Two models were utilized to cross-validate the results. Using the 10 most significant bacterial species from each diet group, we were able to correctly classify the food consumed up to 87% of the time. Cross validation confirmed that our model was not predictive of the trial cohort, but rather the treatment diet. These results reveal promise in accurately predicting consumption of foods that contain nondigestible components (e.g., dietary fiber) using bacterial species as biomarkers. These approaches and subsequent determination of gastrointestinal bacterial biomarkers may be useful fidelity measures to ascertain study compliance. Long term, these approaches may help inform diet-microbiota-tailored recommendations to improve health outcomes.