Fig. 5: Sc-SPORT identifies single-cell RNA structure dynamics.

a, Top: a schematic showing the timelines and cellular identities during neuronal differentiation in this study. Bottom: density plots showing the distribution of gene-level heterogeneity in single cells in each stage from hESC (day 0) to NPC (day 7), iNEU (day 8) and NEU (day 14, top to bottom). Heterogeneity is calculated using the deviation of cosine distance. Only shared genes across four stages were included in the distribution. b, Sankey plots showing the changes in window-level heterogeneity during the neuronal differentiation process. To simplify, shared windows of four stages are grouped into four quartiles according to their heterogeneity values in each stage. c, K-means clustering with dynamic time warping by their heterogeneity across the four stages of neuronal differentiation. We identified six clusters that show different patterns of heterogeneity during differentiation. Left: violin plots show the distribution of structural heterogeneity in the four stages from left to right for each cluster. The bars in the violin plot represent the median and the interquartile range of heterogeneity. The numbers of windows in each cluster are as shown. Right: pie charts show the distribution of the windows present in 5′ UTR, CDS and 3′ UTR of mRNAs. d, Line plots showing per-nucleotide heterogeneity (R²) of 18S rRNA in 76 hESC cells. e, Zoomed-in view showing the location of our identified heterogeneous region in 18S rRNA in its three-dimensional model. The three-dimensional structure is obtained from PDB (PDB ID: 4v6x). We have colored the transfer RNA in green and our changing regions in red. f, A bar plot showing the ∆mutation rates of 18S rRNA between single cells in the S phase and G2/M phase (top). ∆Mutation rates are calculated by subtracting the pseudobulk reactivity of cells in the S phase and the G2/M phase. The black dashed box shows the heterogeneous region (1,590–1,830) and is zoomed-in as below.