Figure 2. Characterization of patterns... | American Society of Hematology

Figure 2.

$Characterization of patterns of injury to microbiota during auto-HCT. (A-C) Microbiota composition of 1161 samples from 534 patients visualized via the t-SNE algorithm. Each point represents a stool sample; those with more similar microbiota composition are clustered together. Dotted lines were manually annotated to highlight features of interest. (A) Higher diversity samples (red and yellow) are clustered away from lower diversity (blue and gray) samples. (B) The same t-SNE projection as in panel A is color coded by day of sample collection. Pretransplant samples (dark purple) cluster in the region of high-diversity samples, whereas samples from days −1 to 20 (light pink and light green) cluster in the low-diversity regions. Of note, some late samples (dark green) again cluster in the early pretransplant or high-diversity region. (C) The same t-SNE projection as in panels A and B is color coded by the most abundant taxon. This shows a cluster of samples with a preponderance of Enterococcus (dark green) in an area of low diversity. Bray-Curtis distances were calculated on operational taxonomic unit (OTU) abundances, and points were color-coded by higher taxonomic ranks as indicated in the color key. f, family; g, genus; o, order; p, phylum. (D) Domination was assessed at the OTU level; bars are color coded at the genus level according to the color key. (E) The fraction of samples with intestinal domination (defined as >30% relative abundance by any single OTU) was >50% by day 7 and was >75% by day 14. (F) Diversity was lower in patients with lymphoma than in those with myeloma or amyloidosis. (G) The same t-SNE projection as in panels A-C is color coded by underlying disease (lymphoma, myeloma, or amyloidosis). Samples did not cluster together significantly by underlying disease.$

Characterization of patterns of injury to microbiota during auto-HCT. (A-C) Microbiota composition of 1161 samples from 534 patients visualized via the t-SNE algorithm. Each point represents a stool sample; those with more similar microbiota composition are clustered together. Dotted lines were manually annotated to highlight features of interest. (A) Higher diversity samples (red and yellow) are clustered away from lower diversity (blue and gray) samples. (B) The same t-SNE projection as in panel A is color coded by day of sample collection. Pretransplant samples (dark purple) cluster in the region of high-diversity samples, whereas samples from days −1 to 20 (light pink and light green) cluster in the low-diversity regions. Of note, some late samples (dark green) again cluster in the early pretransplant or high-diversity region. (C) The same t-SNE projection as in panels A and B is color coded by the most abundant taxon. This shows a cluster of samples with a preponderance of Enterococcus (dark green) in an area of low diversity. Bray-Curtis distances were calculated on operational taxonomic unit (OTU) abundances, and points were color-coded by higher taxonomic ranks as indicated in the color key. f, family; g, genus; o, order; p, phylum. (D) Domination was assessed at the OTU level; bars are color coded at the genus level according to the color key. (E) The fraction of samples with intestinal domination (defined as >30% relative abundance by any single OTU) was >50% by day 7 and was >75% by day 14. (F) Diversity was lower in patients with lymphoma than in those with myeloma or amyloidosis. (G) The same t-SNE projection as in panels A-C is color coded by underlying disease (lymphoma, myeloma, or amyloidosis). Samples did not cluster together significantly by underlying disease.

This Feature Is Available To Subscribers Only