Short- and long-term microbiota dynamics. (A) Study schema. Induction chemotherapy typically starts within a few days after admission and is completed by day 7. Bone marrow aplasia occurs around day 14, which we used to define early (ST1) vs late (ST2) short-term samples. (B) Violin plot shows sample distribution over time. The duration (C) and number (D) of all antibiotics and antibacterial antibiotics during the initial hospitalization. (E) Aitchison distance between each sample and the baseline sample from the same patient. Connected points represent samples from the same patient. (F) Aitchison distance between each sample and the last short-term sample from the same patient. Regression lines for short-term samples are derived from a linear mixed-effect regression, where patient ID was a random effect and sample collection day, measured from the first day of chemotherapy (E) or the last short-term sample (F), was a fixed effect. (G) The top 3 genera containing lost (left panel) and gained (right panel) ASVs between the baseline sample and the last long-term sample for each patient. The y-axis shows the proportion of such ASVs belonging to each genus. Each point represents data from 1 patient. (H) Genus-level relative abundances in baseline (BL) and last long-term (LT) samples from each of the 16 patients with long-term samples. The 5 most abundant genera in each sample were selected, and the combined set of genera generated from all samples was used to plot the stacked bars. Box plots in (C), (D), and (G) show the median (horizontal line), mean (diamond), and interquartile range.