Yen Ting Lin – Computational Biophysics Candidate Seminar

From single-cell experiments to statistical inference of gene expression models: A physicist’s perspective

In this talk, I will first introduce an experimental technique—single-molecule RNA fluorescent in situ hybridization (sm RNA FISH)—which measures transcribed mRNA and the discrete state of activation in a single cell, and provides a “snapshot” of the stochastic process of gene expression. Then, I will discuss how we use a class of coarse-grained stochastic models, formulated as continuous-time and individual-based chemical reactions in a well-mixed environment, to infer kinetic properties of stochastic gene expression from the experimental data. I will present an accurate sampling procedure which is up to 1000-fold speed-up compared to conventional algorithms to efficiently solve the problem numerically. The increased efficiency permits us to go beyond standard fitting procedures and enter to the realm of statistical inference and selection of the most explanatory model from O(2000) candidate models. In the final part of the talk, I will present a high-level description of how we carry out the full-scale Bayesian analysis on our continuous-time probabilistic models using data from discrete-time observations, and how the results were used to inform potential regulatory mechanisms in our experimental systems. The outcome of the analysis, the uncertainty quantification of the parameters and model structures, will be presented.