Skip to main content
. Author manuscript; available in PMC: 2016 Sep 24.
Published in final edited form as: Cell. 2015 Sep 24;163(1):95–107. doi: 10.1016/j.cell.2015.08.059

Figure 3. Diet-discriminatory OTUs are robust to different donor microbiota and motility phenotypes.

Figure 3

(A) Out-of-bag estimated error rates in a Random Forests model for predicting diet, stratified by donor microbiota, as a function of numbers of diet-discriminatory OTUs. For each microbiota, 40 OTUs were sufficient to discriminate diet, yielding a total of 87 unique OTUs across six microbiota donors from five cultural/dietary traditions in the 6-phase travel experiment. (B) Evidence for the robustness of diet-discriminatory OTUs to donor microbiota and motility phenotype. Feature importance scores of 87 diet-discriminatory OTUs in each diet-microbiota context are represented in this heat map. A sparse Random Forests model built using these diet-discriminatory OTUs accurately predicted diet in the 3-phase travel experiment. See also Table S5.