Fig. 7

AGP food differential abundance vectors cluster dietary terms by metabolic content. Each value in the heatmap represents the log-fold-change of each species-term combination, normalized across rows by Z-score. Feature selection on genera was performed by identifying the 20 genera most capable of dividing terms into 2 clusters using the ANOVA F-statistic. The resulting clustering of diet-related terms broadly divides dietary differential abundance vectors into protein-rich, simple carbohydrate, and complex carbohydrate clusters