Extended Data Fig. 4: Random Forest Model Summary for Features Linked to CDI Recurrence.
Random forest model results for microbial, metabolic, and immune features predicting CDI recurrence (at screening) or associated with CDI recurrence (post-screening) in VE303 recipients A, including all participants and B, excluding participants who recurred later than Day 14. The heatmaps represent the frequency with which the feature was selected to be important across 100 model iterations. Features are plotted if permutation importance < 0.1 and frequency > 0.25. Boxes show feature associations with recurrence or non-recurrence identified by the model at the corresponding timepoint.
