Technology Category
Machine Learning: Active LearningReasoning under Uncertainty: Probabilistic InferenceCognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning
Simulation-trained conditional normalizing flows for likelihood approximation: a case study in stress regulation kinetics in yeast
🤖 AI Summary
Conventional Markovian likelihood modeling fails to infer instantaneous gene activity from single-cell snapshots due to history dependence in protein counts—arising from non-exponential cell-division intervals and asymmetric partitioning of proteins at division.
Method: We propose the first simulation-driven, conditional normalizing flow (cNF) method explicitly designed to capture inheritance effects across cell divisions. Our approach jointly models single-cell lineage trajectories, fluorescent-protein inheritance dynamics, and flow-cytometry data.
Contribution/Results: Applied to the yeast *glc3* promoter under nutrient stress, the method achieves high-fidelity inference of transient activation kinetics. We find activation is extremely rare (<5% probability) and short-lived (minute-scale duration), overturning the prior assumption of broad, low-level expression. Moreover, we demonstrate that snapshot-based analyses systematically overestimate gene activity—a widespread bias attributable to unmodeled inheritance dynamics.
Technology Category
Application Category
📝 Abstract
Physics-inspired inference often hinges on the ability to construct a likelihood, or the probability of observing a sequence of data given a model. These likelihoods can be directly maximized for parameter estimation, incorporated into Bayesian frameworks, or even used as loss functions in neural networks. Yet, many models, despite being conceptually simple, lack tractable likelihoods. A notable example arises in estimating protein production from snapshot measurements of actively dividing cells. Here, the challenge stems from cell divisions occurring at non-Exponentially distributed intervals with each division stochastically partitioning protein content between daughter cells, making protein counts in any given cell a function of its full division history. Such history dependence precludes a straightforward likelihood based on a (standard Markovian) master equation. Instead, we employ conditional normalizing flows (a class of neural network models designed to learn probability distributions) to approximate otherwise intractable likelihoods from simulated data. As a case study, we examine activation of the emph{glc3} gene in yeast involved in glycogen synthesis and expressed under nutrient-limiting conditions. We monitor this activity using snapshot fluorescence measurements via flow cytometry, where GFP expression reflects emph{glc3} promoter activity. A na""ive analysis of flow cytometry data ignoring cell division suggests many cells are active with low expression. However, fluorescent proteins persist and can be inherited, so cells may appear active from retaining ancestral fluorescence. Explicitly accounting for the (non-Markovian) effects of cell division reveals emph{glc3} is mostly inactive under stress, showing that while cells occasionally activate it, expression is brief and transient.
Problem
Research questions and friction points this paper is trying to address.
Approximating intractable likelihoods in protein production estimation
Modeling non-Markovian cell division effects on protein counts
Correcting misinterpretation of gene activation in yeast stress response
Innovation
Methods, ideas, or system contributions that make the work stand out.
Conditional normalizing flows for likelihood approximation
Simulation-trained neural networks to learn distributions
Non-Markovian cell division effects explicitly modeled
🔎 Similar Papers
No similar papers found.
