OS and PFS in auto-HCT patients is correlated with perineutrophil engraftment fecal diversity. (A) Histogram of sample collection times relative to transplantation for 1161 samples from 534 unique patients. (B) Fecal diversity, as measured by 16S sequencing and the inverse Simpson index, declined comparably at both institutions during the course of transplantation in 841 samples from 384 unique patients at MSKCC and 207 samples from 121 unique patients at Duke. Each point represents a stool sample color coded by institution. Curves are loess-smoothed averages. The y-axis was square-root scaled using the scale_y_sqrt function of the R package ggplot2. The same data are plotted together with samples through day 100 in supplemental Figure 1. (C) Fecal diversity in early pretransplant samples (defined as the earliest sample available per patient during days −15 to −3) was lower than the median diversity of 313 samples from 212 healthy participants of the HMP and 34 samples from 34 healthy MSKCC volunteers. The data from HMP and healthy MSKCC volunteers have been previously published²⁴  and are reused here as controls. MSKCC includes 70 samples from 70 unique patients and Duke includes 38 samples from 38 unique patients. MSKCC healthy volunteer median was 15.87 (standard error [SE], 1.39), HMP median was 12.05 (SE, 0.50), MSKCC auto-HCT median was 9.39 (SE, 0.67), and Duke median was 9.68 (SE, 1.04). Another pretransplant window is used in supplemental Figure 6. The ends of the whisker lines represent the minimum and maximum values within 1.5× the interquartile range. (D) Patients with higher diversity in available perineutrophil engraftment (days +9 to +16) samples are at a reduced risk of progression or death. Patients were split into above-median (high-diversity) and below-median (low-diversity) groups in each of these day +17 landmark analyses by using the median inverse Simpson index (4.03) as the cutoff value. In all, 181 MSKCC and 59 Duke patients’ samples were analyzed. Analyses of diversity as a continuous variable and multivariable models are provided in supplemental Table 4 and analyses of pretransplant therapies are provided in supplemental Figure 6.