K Hey, H Momiji, K Featherstone, J Davis, M White, D Rand and B Finkenstadt
Inference for a transcriptional stochastic switch model from single cell imaging data
Abstract: Stochastic Reaction Networks (SRNs) can be used to model the temporal behaviour of gene regulation in single cells. In particular, SRNs can capture the features of intrinsic variability arising from the intracellular biochemical processes. However, inference for SRNs is
computationally demanding due to the intractability of the transition densities. This paper will show how state space models provide a unifying framework for approximating SRNs with particular attention given to the linear noise approximation (LNA) and an alternative model specific approximation. This methodology has been applied to single cell imaging data measuring expression levels of the human prolactin gene. Transcription is modelled by a random step function relating to bursts in transcriptional activity and we will demonstrate how reversible jump MCMC can be used to infer the switching regimes of this gene within single cells of mammalian tissue. Bayesian Hierarchical; Linear Noise Approximation; Particle Gibbs; Reversible Jump MCMC; Stochastic Reaction Networks; State Space Models; Transcription.