Fig. 2

Differentiated clones from hiPS precursors sometimes cluster in expression state following cardiac differentiation. a Schematic of barcoding approach for labeling hiPS cells prior to cardiac differentiation. We transduced 4000–8000 hiPS cells at an MOI of ~0.03 with the Rewind [47] barcode library. After 3 days (3–4 population doublings), we seeded cells for cardiac differentiation. On day 14 of cardiac differentiation, we performed single-cell RNA sequencing. We asked whether cells sharing a barcode were more transcriptionally similar on day 14 than randomly sampled barcoded cells. b Table of all 49 barcode clones comprising 20 or more cells within a differentiation, ranked by homogeneity of expression states of constituent cells, as measured using the “Jensen-Shannon distance”, i.e., the observed Jensen-Shannon distance between the cluster probability distribution associated with each barcode and a random cluster probability distribution generated from 1000 samples of a matched number of cells. “Random mean +/− SD” refers to the mean and standard deviation of randomized Jensen-Shannon distances (see “Methods”), which were used to calculate the “p-value” column. “Cells” refers to the number of constituent cells per barcode. Pullout plots for 4 of the profiled barcodes (teal background) labeling 30–40 constituent cells, exhibiting a range of Jensen-Shannon distances (suggesting Jensen-Shannon distance is not a product of cell number) are described as follows. Maintaining the organization provided by UMAP, we plotted all cells in the analysis (gray) and recolored cells corresponding to each featured barcode in teal. We also plotted bar graphs for observed cluster probability distribution (teal) and the average random cluster probability distribution (gray). Finally, we plotted histograms demonstrating the distribution of randomized Jensen-Shannon distances between random cluster probability distributions generated from 1000 additional random samples and the average random cluster probability distribution described above (gray) versus the observed Jensen-Shannon distance between the featured barcode cluster probability and the average random cluster probability distribution (teal vertical line)