Fig. 8: Prediction performance between different biotic levels in different sampling scales.

The comparison of cross-validation results (a–c) and prediction accuracies (d–f) between different biotic levels in different sampling scales. a–c Scatter plots showing the predicted and observed values of relative abundances of different biotic levels inferred from dietary compositions. Colors represented different biotic levels (dark gray: phylum; light gray: family; black: ZOTU). The diagonal line represented perfect prediction (predicted value = observed value). d–f The accuracy of random forest models in predicting relative abundances in different taxa. Significant differences among different datasets were denoted by lettering (P < 0.05, ANOVA followed Duncan’s test). The detailed information was summarized in Supplementary Table 6.