Parameterization of a Stochastic Model of Gene Expression from Adaptive Imaging Cytometry Data

Supplementary material to a manuscript under review. The manuscript can be downloaded from the link below. 

Abstract

Recent decades have yielded an understanding of cellular behavior as arising from a complicated soup of noisy molecular interactions. Systems biologists use this perspective to build models to explain cellular behavior, while synthetic biologists attempt to forward engineer novel function in this complex environment. It is well known that some cellular behaviors can be explained only when the stochasticity of the underlying molecular dynamics are considered; however, models that capture this molecular noise are exceedingly difficult to construct. First, current datasets typically do not capture time courses of individual cells and rely on fluorescent reporters that do not detail the dynamics of underlying components, such as mRNA. Second, matching stochastic models to single-cell data is far more difficult than matching deterministic models to population averages. In this work, we addressed both of these concerns by using a novel instrument based on time-lapse microscopy and by applying a distribution-based method to assess the match between model and data. We demonstrate our model’s ability to match our experimental data in detail, and then use the model to successfully predict the behavior of a modified system. Our approach should lead to more predictive models of simple genetic systems in both systems and synthetic biology.

本材料为待审手稿的补充资料。该手稿可通过下述链接下载。

摘要

近数十年来，学界已达成共识：细胞行为源于充斥着噪声分子相互作用的复杂体系。系统生物学家以此视角构建模型以阐释细胞行为，而合成生物学家则尝试在此复杂环境中正向工程化构建全新功能。众所周知，部分细胞行为仅在考虑底层分子动力学的随机性时才能得到合理解释；但捕捉这类分子噪声的模型构建难度极大。
其一，当前主流数据集通常无法获取单个细胞的时间进程数据，且依赖荧光报告基因(fluorescent reporters)，无法揭示诸如信使RNA(mRNA)这类底层组分的动态变化。其二，将随机模型与单细胞数据进行拟合的难度，远高于将确定性模型与群体平均数据进行拟合的难度。
本研究中，我们通过采用基于延时显微镜(time-lapse microscopy)的新型实验装置，以及运用基于分布的方法评估模型与数据的拟合程度，解决了上述两大问题。我们详细验证了模型与实验数据的拟合能力，并利用该模型成功预测了改造后体系的行为。本研究方法有望推动系统生物学与合成生物学领域中，简单遗传系统的可预测模型构建工作。