Inference of Lineage-Resolved Cell Identities in Uncompressed C. elegans Embryos

Lineage-based cell identification in C. elegans typically requires intensive live imaging and tracking. Recent methods attempt to automate cell identification on the basis of spatiotemporal atlases. We present EmbAlign, an automated 3D registration framework that determines lineage identities from single embryo snapshots. EmbAlign retrieves reference templates from a spatiotemporal atlas and refines assignments using an iterative Sinkhorn alignment procedure, robustly handling positional variability and arbitrary orientations in uncompressed embryos. EmbAlign achieves 96.9% accuracy up to the 190-cell stage and includes a diagnostic layer for continuous scoring (AUPRC = 0.546), converting raw spatial data into lineage aware datasets.