Scalable and flexible inference framework for stochastic dynamic single-cell models.

Area of Science:

• Systems biology
• Single-cell analysis
• Computational biology

Background:

• Understanding dynamic biological processes and cell-to-cell variability is crucial in systems biology.
• Single-cell fluorescent microscopy generates rich data, but extracting mechanistic insights remains challenging.
• Existing inference methods struggle with efficiency, generality, and accessibility for single-cell data.

Purpose of the Study:

• To develop a scalable and flexible framework for Bayesian inference in stochastic dynamic state-space mixed-effects single-cell models.
• To enable the inference of model parameters from single-cell data, accounting for both intrinsic and extrinsic noise.
• To apply the framework to understand cell-to-cell variability in biological pathways.

Main Methods:

• Developed a Bayesian inference framework for state-space mixed-effects single-cell models with stochastic dynamics.
• Incorporated methods for modelling intrinsic noise using stochastic simulators (exact or approximate).
• Accounted for extrinsic noise using time-varying or time-constant parameters that vary between cells.

Main Results:

• Successfully inferred model parameters for complex single-cell models with both intrinsic and extrinsic noise.
• Applied the framework to study the SNF1 nutrient sensing pathway in Saccharomyces cerevisiae.
• Identified hexokinase activity as a source of extrinsic noise and sugar availability as a key determinant of cell-to-cell variability.

Conclusions:

• The developed framework provides an efficient and accessible tool for mechanistic modeling of single-cell data.
• This approach can elucidate the sources of heterogeneity in cellular processes.
• Findings highlight the role of sugar availability and hexokinase activity in yeast nutrient sensing variability.