Trajectory inference from single-cell genomics data with a process time model.

Area of Science:

• Single-cell biology
• Computational biology
• Systems biology

Background:

• Single-cell transcriptomics captures cell states but ordering cells using "pseudotime" lacks physical meaning.
• Existing methods struggle with discrete cell clusters and continuous cell state transitions.

Purpose of the Study:

• To develop a principled modeling approach for inferring "process time" from single-cell data.
• To introduce Chronocell, a computational tool for biophysical trajectory inference.
• To provide a method that bridges trajectory inference and clustering.

Main Methods:

• Formulating cell state transitions using a biophysical model.
• Inferring latent variables representing timing within a biophysical process.
• Implementing the Chronocell model for identifiable parameter inference.

Main Results:

• Chronocell successfully interpolates between continuous and discrete cell states.
• Analysis of diverse datasets revealed distinct cellular distributions along process time.
• Parameter estimates for degradation rates align with metabolic labeling data.

Conclusions:

• Chronocell offers a biophysically meaningful framework for analyzing single-cell dynamics.
• The method aids in assessing dataset suitability and understanding biological timing.
• Accurate process time inference depends on dataset quality and model evaluation.