Extended Data Fig. 2: ATAC-seq data QC.

a, Peak calling versus library sequencing depth. We observed a slower rise in the number of new peaks called after 15 millions filtered read pairs. This indicates a reasonable balance between read depth and number of peaks called using an average of 14 million read pairs after filtering in our samples. b, Insert size histograms for 3 randomly selected neuron and progenitor samples. c, PCA plot for ATAC-seq data (N = 137) before batch correction (left) and after batch correction (right), colored by sorter. We corrected normalized reads within ATAC-seq peaks in neurons by sorter locations. Then, we corrected normalized reads within ATAC-seq peaks in neurons and progenitors by cell culture round. d, Correlations of batch corrected normalized reads across donors and within donors. Correlations within donors was significantly higher than correlations across donors in progenitor (n = 15). Correlations within donors were higher than correlations across donors in neurons (n = 4), but not significant (p = 0.07). P values are estimated by two-sided wilcoxon tests. The center of the box is median of the data, the bounds of the box are 25th percentile and 75th percentile of the data, and the whisker boundary is 1.5 times the IQR. Maximum and minimum are the maximum and minimum of the data. e, Correlations between PC1 to PC10 from normalized reads in neurons with known technical and biological factors. f, Correlations between PC1 to PC10 from batch correction normalized reads in progenitors with known technical and biological factors.