Fig. 5: Quantifying lineage-specific regulation strategies through metabolic labeling.

a, Cells are metabolically labeled in pulse–chase experiments⁹ followed by simultaneous sequencing. Pulse experiments involve incubation with nucleoside analogs for varying durations; in chase experiments, cellular mRNA fully incorporates nucleoside analogs during a long incubation, followed by washing out of these nucleosides for varying durations. b, For each cell, gene and labeling duration, we identify the number of neighbors such that a predefined number of cells with non-trivial counts are included in the neighborhood, illustrated here for an exemplary cell A. These cells are then used to estimate cell and gene-specific transcription and degradation rates α and γ, respectively, to model the dynamics of labeled mRNA. c, UMAP embedding highlighting terminal states identified using CellRank 2 and dynamo. Green ticks indicate that a method recovered the corresponding terminal state, and red crosses indicate that the terminal state was not identified. d, Ranking of drivers for each lineage identified by different methods. Dynamo identified only enterocytes as terminal and, thus, provides a gene ranking only for this lineage. For dynamo, the mean gene ranking and corresponding 95% confidence band are shown (Methods). e, Inferred transcription (left) and degradation (right) rates of top-ranked known drivers of the goblet lineage. f, Same as e, but along the enterocyte lineage.