New Model Reconstructs Cell State Dynamics

A research team from Kyushu University has developed a computational method named ddHodge that reconstructs how cells transition and decide their fates. Published in Nature Communications, the study presents a geometry-preserving framework that enhances understanding of developmental, regenerative, and disease processes by describing how cells move through different states.

In biology and medicine, a central challenge is understanding how a cell determines its identity—for instance, becoming a nerve or muscle cell. Researchers often use single-cell RNA sequencing (scRNA-seq) to identify which genes are active in individual cells. However, this technique captures only static snapshots, not the dynamic evolution of cell states. Methods like RNA velocity address this limitation by estimating a cell’s direction and speed of change but simplify complex gene data into fewer dimensions. This compression causes loss of structural details essential for distinguishing cells that are stable from those in transition.

To address this, Kazumitsu Maehara and Yasuyuki Ohkawa developed ddHodge, a method that accurately reconstructs cellular dynamics while preserving geometric information. “My background is in statistical science, and during my graduate training, I was exposed to HodgeRank, a method used in ranking problems such as PageRank,” Maehara explained. “When I later moved into life-science research, I realized that the same mathematical idea could help interpret the complex, high-dimensional transitions in single-cell data.”

The technique uses Hodge decomposition, a mathematical principle that decomposes motion across a landscape of possible cell states into three measurable components. The gradient represents the overall direction of change, while the curl and harmonic components describe cyclical or repeating processes, including the cell cycle. This approach preserves the geometric relationships within high-dimensional datasets, maintaining information usually lost during dimensionality reduction.

When applied to single-cell data from approximately 46,000 mouse embryonic cells, ddHodge revealed that more than 88% of gene expression dynamics during early development were explained by the gradient component. This observation supports the view that cells progress toward stable states and diverge from unstable branching points. The analysis also identified key genes that affect the stability of cell states and lineage commitments. Simulations confirmed that ddHodge could reconstruct these dynamics with about 100 times greater accuracy than standard methods, even with partial or noisy data. 

Maehara noted that “ddHodge can quantitatively describe, within a high-dimensional space, in which direction, how fast, and how stably cells change. We expect it to contribute broadly to understanding diverse biological phenomena, including embryonic development, tissue regeneration, and cancer progression.”