a, Summary of bioinformatics workflow. b, Cell doublets are overlaid on t-SNE graphs using DoubletFinder. n = 8,710 and n = 6,644 single-cell transcriptomes for whole epithelium (samples 04 and 06) and enriched crypts (samples 05 and 07), respectively; see Methods. c, d, Unsupervised clustering of scRNA-seq data from whole epithelia (c; n = 8,710 single-cell transcriptomes) and isolated crypts (d; n = 6,644 single-cell transcriptomes). Unsupervised clusters are overlaid on the t-SNE map and are indicated by different colours and labelled according to their identity, on the basis of expression of cell-type-specific marker-gene expression. Cells from untreated controls (blue dots) or irradiated intestines (red dots) are also plotted as indicated. In c, median expression of epithelial cell markers (bottom left) and lymphocyte markers (bottom middle) are also shown. Heat maps (right) showing expression of indicated marker genes (c (right), d (right)) that identify distinct cell-types are shown. Note that the cluster 18 in d, highlighted by the question mark, does not express any lineage-specific markers. Gene-expression values are shown as median of log10(counts per million (CPM)). IM, immune cells.