Reconstruction of cell diversity and cell lineages from somatic mutations in single-cell transcriptomic data

Abstract

Abstract Single-cell RNA sequencing (scRNA-seq) enables high-resolution profiling of individual cells. However, inferring temporal relationships among cells remains a challenge. Here, we present real-time course analysis (RTCA), a simple direct phylogenetic signal framework that reconstructs cell lineages from somatic variants detected across cells using variants derived from nuclear-encoded transcripts in scRNA-seq data. Our simulations demonstrate that only a modest increase in informative variant sites is sufficient to maintain accurate tree reconstruction, even with a tenfold increase in the number of cells. This scalability makes RTCA applicable to datasets generated by recent single-cell sequencing technologies and is also suitable for reanalyzing existing datasets. We also performed a comparative analysis between our method and a representative genotype-mediated inference framework, PhylinSic, which shares certain conceptual similarities with RTCA. The simulation results show that RTCA is more robust to sparse mutation signals and dropout-induced missing data than PhylinSic. In an application to the datasets from 2 healthy human placental samples, RTCA successfully reconstructed bifurcating phylogenetic trees. By mapping expression-based cell type clusters onto the trees, we evaluated the degree of monophyly within lineages and found our results consistent with known placental differentiation pathways. The identified cell lineages also aligned with classifications based on gene expression and pseudotime analysis. Compared to gene expression-based and pseudotime analysis, RTCA provides a temporally based model of cell trajectories, integrating lineage and expression information in a biologically meaningful manner. In summary, RTCA offers a scalable, cost-effective solution for reconstructing developmental processes in complex normal tissues.