Quantitative model analysis with diverse biological data: Applications in developmental pattern formation
Quantitative model analysis with diverse biological data: Applications in developmental pattern formation
复制标题
DOI：
10.1016/j.ymeth.2013.03.024
复制
发表时间：
2013-07-15
期刊：
影响因子：
4.8
通讯作者：
Umulis，David M.
中科院分区：
文献类型：
--
作者：
Pargett，Michael;Umulis，David M.
转录因子和信号网络的数学建模被广泛用于理解机制是否以及如何工作，并推断产生与观察数据一致的模型的调节相互作用。这两种建模方法都是由实验数据提供信息的，然而，大部分可用或甚至可获得的数据都不是定量的。非严格定量的数据不能用于经典的、定量的、基于模型的分析，这些分析测量测量的观测值与该观测值的模型预测值之间的差异。为了弥合模型与数据之间的差距，已经开发了各种技术来测量模型“适应性”，并提供随后可用于模型优化或模型推断研究的数值。在这里，我们讨论了一系列传统和新颖的技术来转换不同质量的数据，并与数学模型进行定量比较。这篇综述旨在告知这些模型分析方法的使用，重点是参数估计，并帮助指导根据可用数据类型选择用于给定研究的方法。应用诸如归一化或最优缩放的技术可以显著提高当前生物数据在基于模型的研究中的效用，并且允许不同类型的数据之间的更大整合。
Mathematical modeling of transcription factor and signaling networks is widely used to understand if and how a mechanism works, and to infer regulatory interactions that produce a model consistent with the observed data. Both of these approaches to modeling are informed by experimental data, however, much of the data available or even acquirable are not quantitative. Data that is not strictly quantitative cannot be used by classical, quantitative, model-based analyses that measure a difference between the measured observation and the model prediction for that observation. To bridge the model-to-data gap, a variety of techniques have been developed to measure model “fitness” and provide numerical values that can subsequently be used in model optimization or model inference studies. Here, we discuss a selection of traditional and novel techniques to transform data of varied quality and enable quantitative comparison with mathematical models. This review is intended to both inform the use of these model analysis methods, focused on parameter estimation, and to help guide the choice of method to use for a given study based on the type of data available. Applying techniques such as normalization or optimal scaling may significantly improve the utility of current biological data in model-based study and allow greater integration between disparate types of data.