Quantifying the Single‐Cell Morphological Landscape of Cellular Transdifferentiation through Force Field Reconstruction

Advancements in sequencing technologies have reshaped our understanding of cell behaviors through transcriptomics, but a gap remains in developing quantitative models for multi-omic data, especially for cellular morphology changes. A pivotal challenge is the lack of cell-specific velocity information, crucial for reconstructing global velocity fields to analyze dynamics and thermodynamics in cellular process. In this study, fibroblast-to-neuron transdifferentiation snapshots are captured and a novel machine learning approach to reconstruct the underlying force field of the morphological change from the sparse sampled single-cell imaging data is developed. The methodology involves decomposing the driving force field into a flow flux force field and a gradient of a time-dependent potential, extending the landscape and flux framework to non-steady-state conditions. This study demonstrates that the reconstructed force field accurately captures the intrinsic morphological landscape of cell fate switching and reveals the impact of noise on state transitions. This approach offers a general framework for analyzing single-cell morphological data and holds promise for application to other single-cell multi-omic datasets lacking inherent velocity information.