Extended Data Figure 1: Summary of read distributions across the three sampled time points.

a, log₁₀ counts of sci-ATAC-seq reads per barcode at each time point are bimodally distributed. A threshold of 500 reads was used to identify barcodes corresponding to valid cells versus background. b, Fragment size distribution at each time point is consistent with expected nucleosomal banding pattern of standard (bulk) ATAC-seq experiments. c, Violin plot for distribution of unique, mappable reads per cell at each time point (2–4 h, n = 8,024; 6–8 h, n = 7,880; 10–12 h, n = 7,181) plotted on a logarithmic scale. White point indicates median value, thick black line extends to 25th and 75th percentile, and thin black lines extend to most extreme values within 1.5 times the interquartile range of the median. The filled colour width represents a density estimate of the distribution of cells along the y axis. d, Fraction of previously characterized DHS covered in at least 10 cells upon sampling a given number of cells (solid lines) as compared to random genomic windows (dashed lines). e, An UpSet plot shows the degree to which the top 20,000 windows overlap between the three time points. Each bar shows the number of sites included in a specific intersection and the ‘peg board’ below shows which particular comparison is included in that bar. f, Bar plot of the number of sites identified as significantly open in each clade (1% FDR; grey bar, first cutoff) and the number of sites specific to that clade (orange bar, second cutoff). Overlaid on the barplot (purple points) is the fraction of sites passing the first cut-off that also pass the second cut-off (count of orange bar/count of grey bar).