Diffusion pseudotime (DPT) is a robust method for reconstructing lineage branching in single-cell data. It measures progression through branching lineages using a random-walk-based distance in diffusion map space and allows for branching and pseudotime analysis on large-scale RNA-seq datasets. Unlike other pseudotime methods, DPT does not involve dimension reduction and accounts for random walks on all length scales, making it more accurate in capturing the underlying temporal order of cellular processes. Single-cell gene expression profiles encode the intrinsic latent temporal order of differentiating cells. DPT reconstructs this order by measuring transitions between cells using diffusion-like random walks, enabling the identification of cells undergoing branching decisions or in metastable states. This approach is particularly useful for identifying critical branching points and quiescent or metastable cells, which lack a clear temporal ordering. DPT is computed in three steps: first, a transition matrix T is determined to approximate dynamic transitions of cells through differentiation stages. Second, the distance dpt(x,y) between two cells is calculated using accumulated transition probabilities. Third, branching points are identified by comparing two random walks starting from the root cell and its maximally distant cell, measuring pseudotimes with respect to these points. DPT was tested on real data, including single-cell qPCR data focusing on early blood development and droplet-based scRNA-seq experiments. It successfully identified branching points and metastable states, revealing distinct developmental trajectories and gene expression patterns. For example, in the blood development dataset, DPT identified three branches corresponding to precursor, blood, and endothelial lineages, with characteristic gene expression patterns in each. DPT outperformed existing methods like Wanderlust and Monocle in terms of accuracy and robustness, particularly in large datasets. It was able to handle the asynchrony of developmental stages and identify key decision genes, showing that metastable states contain more genes than developmental stages, with less bimodal expression. DPT is statistically robust and can be scaled to large datasets without dimension reduction. It provides a powerful and reliable tool for ordering cells according to their state on differentiation trajectories in single-cell transcriptomics studies. By capturing the underlying temporal order of cellular processes, DPT enables the inference of regulatory relationships with higher confidence than based on perturbations alone.Diffusion pseudotime (DPT) is a robust method for reconstructing lineage branching in single-cell data. It measures progression through branching lineages using a random-walk-based distance in diffusion map space and allows for branching and pseudotime analysis on large-scale RNA-seq datasets. Unlike other pseudotime methods, DPT does not involve dimension reduction and accounts for random walks on all length scales, making it more accurate in capturing the underlying temporal order of cellular processes. Single-cell gene expression profiles encode the intrinsic latent temporal order of differentiating cells. DPT reconstructs this order by measuring transitions between cells using diffusion-like random walks, enabling the identification of cells undergoing branching decisions or in metastable states. This approach is particularly useful for identifying critical branching points and quiescent or metastable cells, which lack a clear temporal ordering. DPT is computed in three steps: first, a transition matrix T is determined to approximate dynamic transitions of cells through differentiation stages. Second, the distance dpt(x,y) between two cells is calculated using accumulated transition probabilities. Third, branching points are identified by comparing two random walks starting from the root cell and its maximally distant cell, measuring pseudotimes with respect to these points. DPT was tested on real data, including single-cell qPCR data focusing on early blood development and droplet-based scRNA-seq experiments. It successfully identified branching points and metastable states, revealing distinct developmental trajectories and gene expression patterns. For example, in the blood development dataset, DPT identified three branches corresponding to precursor, blood, and endothelial lineages, with characteristic gene expression patterns in each. DPT outperformed existing methods like Wanderlust and Monocle in terms of accuracy and robustness, particularly in large datasets. It was able to handle the asynchrony of developmental stages and identify key decision genes, showing that metastable states contain more genes than developmental stages, with less bimodal expression. DPT is statistically robust and can be scaled to large datasets without dimension reduction. It provides a powerful and reliable tool for ordering cells according to their state on differentiation trajectories in single-cell transcriptomics studies. By capturing the underlying temporal order of cellular processes, DPT enables the inference of regulatory relationships with higher confidence than based on perturbations alone.