Figure 4.

Comparison of prediction accuracies between the FCP and MAP methods

A. Prediction accuracies of the FCP and MAP methods in AMD samples. The prediction accuracies for the training dataset and validating dataset in AMD samples are measured using Bray–Curtis similarity, with the average accuracies also shown. B. Cross-validation of the predicted relative microbial abundances with MAP and FCP methods in AMD communities. The linear regression of the FCP model is expressed as y = 0.96x + 0.003 (R² = 0.92) and that of the MAP model is expressed as y = 0.79x + 0.03 (R² = 0.72), respectively. C. Relative influence of environmental factors on AMD biofilms using GBM, FCP, and MAP methods. D. Comparison of prediction accuracies between the FCP and MAP models in human gut microbiota samples. The accuracies are measured using Bray–Curtis similarity. FCP, functional gene-based composition prediction; MAP, microbial assemblage prediction.