Reconstructing Complex Cell State Transitions with Real-Time Inference

Cell state transitions arise from coordinated and dynamic changes in gene expression. While pseudotime algorithms have enabled the reconstruction of such dynamics from static single-cell transcriptomic data, they typically lack biologically meaningful time scaling. Here, we present RETRO, a computational algorithm that infers realistic temporal trajectories of cell state transitions by integrating single-cell gene expression profiles with low-resolution temporal constraints. RETRO accurately reconstructs complex, branching, and multi-step transition patterns while providing inferred time courses on a proper biological timescale. Applied to both synthetic datasets and experimental systems including hematopoiesis and fibroblast reprogramming, RETRO outperforms existing approaches in capturing the temporal order and dynamics of cellular transitions. This approach enables mechanistic insights into cell fate decision processes and facilitates systems biology modeling.