Fig. 3: Inferring single-cell age reversal in iPSC induction with EpiTrace.

a, Schematic overview of the in vitro chemical induction of human pluripotent stem cells (‘Primed’) back to 8-cell like cell (8CLC) state, through serially culturing in 4-cell-like medium (4CL, three passages (P3)) and the enhanced 4CL-medium (e4CL). b, DNAm age of D0 (day 0, Primed) and D12 (day 12, 4CL) cultures and sorted 8CLCs from D17 (day 17) culture, from WGBS data. n = 2 independent biological repeats in each group. c, Single-cell age estimated with EpiTrace from the D17 scMultiomic dataset. Sample numbers of biologically independent cells: n = 483 (primed); 33 (interim); and 61 (8CLC). d, Correlation of inferred age from DNAm (whiskers denote minimum/maximum, central point denotes median value, y axis) or single-cell EpiTrace age (whiskers denote 25%/75%, central point denotes median value, x axis) from the same set of cells. Correlation R and P value: Pearson’s. Sample numbers of WGBS and single cells were as in b and c. e, UMAP of scMultiomic-sequenced D17 culture with single-cell evolution trajectories built with kernels combining EpiTrace age and RNA velocity information. f, Schematic overview of the in vitro chemical induction of human adult fibroblasts toward chemically induced pluripotent stem cells (CiPSC). Both the uninduced and intermediate stage II cultures were sequenced by scATAC. 5-azaC, 5-azacytidine; C6NYSA, combination of CHIR99021, 616452, TTNPB, Y27632, SAG and ABT869; hADSC, human adipose stromal cell (mesenchymal stromal cell); HEF, human embryonic fibroblast; JNKIN8, c-Jun N-terminal kinase inhibitor; T5J, tranylcypromine, JNKIN8 and 5-azaC. g, Single-cell age estimated with EpiTrace from f. The induced cultures were either subjected to the full induction paradigm (+Chem: C6NYSA + T5J) or had 5-azaC or JNKin8 removed (−5aza, −JNKin8). Sample numbers of biologically independent cells: n = 8,826 (uninduced); 4,667 (+Chem, −JNKin8); 10,257 (+Chem, −5aza); and 8,671 (+Chem stage II). Statistical comparisons are shown between groups by two-sided Wilcoxon test. h, Correlation coefficient between the ChrAcc on each ATAC peak and EpiTrace age estimated from the MSC experiment (x axis) or the FB experiment (y axis). Peaks of interest are labeled, colored by their genomic ___location class. i, Prediction of sample age by DNAm from WGBS data of chemical induction of iPSCs. Chemical reprogramming induces genome-wide demethylation and an increase in DNAm age, as reported previously³¹, whereas the addition of 5-azaC globally reduces DNAm to increase DNAm age. Removal of 5-azaC blocks DNAm age from increasing. Sample numbers of biologically independent samples: n = 4 (uninduced); 2 (C6NYSA); 1 (−JNKIN8); 1 (−5azaC); 2 (C6NYSA + T5J); and 4 (iPSC/hESC). j, Scatter plot of WGBS DNAm age (x axis) and mean single-cell EpiTrace age (y axis) of the same sample. For box plots, the upper and lower bounds of boxes show 25% and 75% percentiles of the data. The median of data is shown as the horizontal line in the box. The distribution minima and maxima, defined as farthest data point distanced ≤1.5 IQR from the box bounds, are shown by the whiskers. The violin plot shows the empirically estimated density distribution of data. Corr.coef., correlation coefficient.