VOUS: Variational Ornstein-Uhlenbeck Stochastics Linking Single-Cell Lineage Tracing with Dynamic Gene Expression

Single-cell gene expression evolves dynamically along cell division histories. However, most existing single-cell methods treat cells as static snapshots, neglecting the rich information encoded in their underlying lineage structures. Recent advances in single-cell lineage tracing now enable the reconstruction of high-resolution lineage phylogenies, providing a natural framework pinpoint exactly when and where transcriptional changes occur. This capability is fundamental to decoding the dynamics of development, differentiation, and disease progression. To fully leverage this lineage information, we present VOUS (Variational Ornstein-Uhlenbeck Stochastics), a flexible probabilistic framework that models stochastic single-cell gene expression over inferred cell lineage trees. By grounding gene expression analysis in explicit cell lineage phylogenies with topology and branch lengths, VOUS enables the inference of continuous expression dynamics, despite the high sparsity and low coverage of sequencing data. We applied VOUS to scRNA-seq data from metastatic lung cancers, identifying gene programs associated with metastasis and potential therapeutic targets. By providing a rigorous foundation for modeling sparse count data on latent tree structures, VOUS establishes a generalizable framework that naturally extends to multi-gene programs, lineage uncertainty, and multi-modal integration, paving the way for a comprehensive atlas of single-cell stochastic dynamics.