arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1403.8057q-bio.QMstat.CO
收藏

具有可测变异性的随机生物系统的有效参数推断

Efficient parametric inference for stochastic biological systems with measured variability

Iain G. Johnston

摘要:

生物学中的随机系统通常在细胞内和细胞之间表现出很大的变异性。这种可变性,以及具有戏剧性的功能后果,提供了有关系统行为的基本细节的信息。这往往是可取的,以推断属性的参数管理这样的系统给定的实验观测的平均值和方差的观测量。在某些情况下,这些观察的可能性的分析形式允许非常有效的推理:我们提出这些形式,并展示其用法。当似然函数不可用或难以计算,我们表明,实施近似贝叶斯计算(ABC)是一个强大的工具,在这些系统中的参数推断。然而,将ABC应用于这些系统所需的计算也可能在计算上是昂贵的,依赖于重复的随机模拟。我们提出了一种ABC方法,廉价地消除了参数空间的不重要区域,通过解决计算简单的平均行为,然后再明确地模拟计算要求更高的方差行为。我们表明,这种方法导致速度大幅提高时,应用于合成和实验数据集。

英文摘要:

Stochastic systems in biology often exhibit substantial variability within and between cells. This variability, as well as having dramatic functional consequences, provides information about the underlying details of the system's behaviour. It is often desirable to infer properties of the parameters governing such systems given experimental observations of the mean and variance of observed quantities. In some circumstances, analytic forms for the likelihood of these observations allow very efficient inference: we present these forms and demonstrate their usage. When likelihood functions are unavailable or difficult to calculate, we show that an implementation of approximate Bayesian computation (ABC) is a powerful tool for parametric inference in these systems. However, the calculations required to apply ABC to these systems can also be computationally expensive, relying on repeated stochastic simulations. We propose an ABC approach that cheaply eliminates unimportant regions of parameter space, by addressing computationally simple mean behaviour before explicitly simulating the more computationally demanding variance behaviour. We show that this approach leads to a substantial increase in speed when applied to synthetic and experimental datasets.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑