Inferring EMT dynamics from cell cycle profiles using a hidden Markov framework

The Epithelial-to-Mesenchymal Transition (EMT) which is a driver of metastasis and a contributor to wound healing, embryogenesis, and trophoblast differentiation is highly context specific, making it difficult to infer robustly from sequencing data. Existing methods often require that epithelial, hybrid, and mesenchymal states all be present in the same dataset and rely on simplifying assumptions that limit generalizability. By contrast, cell cycle stage is routinely and reliably inferred from transcriptomic profiles. We leverage this robustness to recover EMT dynamics when direct EMT inference fails. Specifically, we learn an emission model that links latent EMT states to observed cell cycle stage frequencies using datasets with known EMT trajectories. We then use this model, together with the initial EMT distribution of a cell line, to demonstrate that we can reconstruct EMT trajectories from time-series measurements of cell cycle stage alone. Finally, we provide an open-source R package and a companion web application to make the approach accessible and reproducible. SIGNIFICANCEElucidating EMT dynamics is central to understanding metastasis and therapy response, yet EMT is notoriously context-specific and often cannot be inferred reliably from transcriptomic data alone. We introduce a population-level hidden Markov model that links latent EMT states to observed cell cycle compositions via an estimated emission matrix thereby enabling the reconstruction of EMT trajectories and transition rates using only cell cycle fractions and a baseline EMT distribution (specific to the cell line). This leverages the strong coupling we observe between EMT progression and cell cycle regulation across cell lines, while remaining robust when direct EMT inference fails. We provide an open-source R package and web app to make these analyses accessible.