Abstract
Cellular development has traditionally been described as a series of transitions between discrete cell states, such as the sequence of double negative, double positive and single positive stages in T-cell development. Recent advances in single cell transcriptomics suggest an alternative description of development, in which cells follow continuous transcriptomic trajectories. A cell’s state along such a trajectory can be captured with pseudotemporal ordering, which however is not able to predict development of the system in real time. We present pseudodynamics, a mathematical framework that integrates time-series and genetic knock-out information with such transcriptome-based descriptions in order to describe and analyze the real-time evolution of the system. Pseudodynamics models the distribution of a cell population across a continuous cell state coordinate over time based on a stochastic differential equation along developmental trajectories and random switching between trajectories in branching regions. To illustrate feasibility, we use pseudodynamics to estimate cell-state-dependent growth and differentiation of thymic T-cell development. The model approximates a developmental potential function (Waddington’s landscape) and suggests that thymic T-cell development is biphasic and not strictly deterministic before beta-selection. Pseudodynamics generalizes classical discrete population models to continuous states and thus opens possibilities such as probabilistic model selection to single cell genomics.
Introduction
In single-cell transcriptomics, "pseudotime" is a standard tool for ordering cells along developmental paths based on genetic similarities. However, pseudotime is fundamentally a coordinate of state, not time. It describes the sequence of events but fails to capture the actual kinetics of development—how long transitions take, the rates of cell division, or the impact of apoptosis. To bridge this gap, we introduce Pseudodynamics, a mathematical framework that integrates real-time sampling data and genetic perturbations to turn static snapshots into a predictive movie of cell population evolution.
The Pseudodynamics Framework
Pseudodynamics shifts from treating development as jumps between discrete stages to modeling it as a continuous density evolving via a reaction-diffusion-advection equation. This partial differential equation (PDE) describes population density over a one-dimensional pseudotime coordinate using three key parameters:
- Advection (Drift): Represents the directed "speed" of development.
- Diffusion: Accounts for stochasticity or "noise" in the developmental process.
- Reaction (Birth-Death): Models population growth or shrinkage due to proliferation and selection.
By maximizing the likelihood of observed data across multiple real-time points, the model reconstructs a quantitative approximation of Waddington’s developmental landscape.
Key Insights in T-cell Development
We applied Pseudodynamics to a newly generated Drop-seq dataset of thymic T-cell development in mice, spanning eight time points from embryonic day E12.5 to birth. The model revealed that T-cell maturation is biphasic, characterized by distinct phases "towards" and "away from" the critical beta-selection checkpoint.
Our analysis identified beta-selection as a "saddle point" in the transcriptomic landscape. At this coordinate, developmental drift drops to near zero, and the process becomes diffusion-dominated, suggesting development is less deterministic and more stochastic during this phase. Following beta-selection, the model captures a rapid expansion in population density, aligning with the known proliferation of cells that successfully rearrange their T-cell receptors.
Validation and Checkpoint Mapping
The framework was further validated using Rag1 and Rag2 knock-out mice, which experience developmental arrest during maturation. Pseudodynamics precisely pinpointed the transcriptomic coordinate of this arrest at a pseudotime of 0.27, matching known biological markers. The delay in development was statistically significant ($p < 10^{-9.6}$), and leave-one-time-point-out cross-validation confirmed that the model accurately "fills in the gaps" for unobserved time points.
Conclusion
Pseudodynamics generalizes classical discrete population models to continuous states, enabling probabilistic model selection in single-cell genomics. While currently optimized for 1D trajectories, future extensions to higher-dimensional state spaces and integration with lineage tracing will further refine our ability to map the speed and selection pressures of cellular maturation.
Summary updated