Spatiotemporal Multiplexing of Single-Cell RNA Velocity Trajectories via Graph-Regularized Optimal Transport for Lineage Deconvolution in Murine Hematopoiesis

Abstract

Reconstructing developmental lineage hierarchies from single-cell transcriptomic data remains a fundamental challenge in systems biology, particularly when asynchronous cell-state transitions obscure pseudotemporal ordering. Here, we present SCRVOT, a graph-regularized optimal transport framework that integrates RNA velocity vectors with spatially resolved transcriptomic anchors to resolve hematopoietic progenitor bifurcations at single-cell resolution. Applied to murine bone marrow datasets comprising 84,000 cells across five developmental timepoints, SCRVOT demonstrated superior branch-point disambiguation compared to existing trajectory inference tools, achieving a 31% improvement in lineage assignment accuracy as measured by clonal barcoding ground truth. The framework further incorporates a Wasserstein-penalized regularization term to mitigate dropout-induced trajectory distortion. These findings establish SCRVOT as a robust computational strategy for dissecting multipotent progenitor fate decisions with direct translational relevance to leukemogenesis modeling.