Spherical PCA: a transparent and scalable approach to quantifying the Waddington Landscape in single-cell RNA-seq data 3144

Abstract Description Understanding cell fate decisions in developmental landscapes is a key challenge in single-cell biology. We present spherical principal component analysis (PCA), an interpretable method for visualizing and quantifying these trajectories. By projecting data onto a unit sphere, spherical PCA preserves global variance, providing a more accurate representation of cell differentiation than deep learning methods, which often distort structure. Applied to the C. elegans cell atlas, spherical PCA revealed “stripe-like” structures aligned with embryonic time, a pattern consistent across additional datasets like mouse embryo and Planarian cells. It improved temporal correlation by 40% over variational autoencoders (Pearson correlation 0.77). Gene analysis highlighted that central housekeeping genes shape the sphere’s core, while lineage-specific genes drive stripe formation. Spherical PCA effectively captures global and hierarchical cellular patterns, offering a computationally efficient tool for single-cell RNA-seq and spatial transcriptomics analysis. This method enhances trajectory analysis and paves the way for new clustering approaches, advancing our understanding of development and disease. Funding Sources This study was supported by The Mark Foundation for Cancer Research and the Bloomberg∼Kimmel Institute for Cancer Immunotherapy at Johns Hopkins University SOM Topic Categories Computational and Systems Immunology (COMP)