Integrated Bayesian Deterministic and Probabilistic Modeling of Phenotype-Genotype Networks in the Metabolic System

该项目为数学建模开发了一种全面的贝叶斯方法，该方法集成了不同的数据类型，考虑了模型的不确定性，并结合了有关生物系统的现有先验知识。 贝叶斯模型整合了两种成熟的建模范式：（1）细胞代谢的确定性数学模型和（2）表型和基因型的因果图模型。 这些建模范式被广泛使用，但遭受重大的限制：确定性模型不反映个体的变化，导致从等位基因变化的酶的通量，和图形建模技术目前忽略了所有的生化途径的先验信息。 该项目通过整合跨越基因组、转录组、代谢组和临床表型的不同形式的数据和变量来克服这些限制。 应用集中在乳腺癌和皮肤癌细胞代谢的基因组规模的代谢重建，以寻找新的分子靶向治疗，副作用，并解开疾病背后的复杂机制。 这一方法将在一个公开提供的软件包中实施，该软件包是为数学和统计学基础最低的用户设计的。 了解遗传学如何影响代谢和调节功能对于疾病的预防和治疗非常重要。 遗传学部分解释了为什么有些人对药物治疗更敏感，而另一些人则没有反应或可能会出现致命的副作用。 高通量表型分析技术的发展使得大规模的分子性状测量成为可能。 数学模型被广泛用于从高通量数据中阐明分子性状网络，并已成为系统生物学的重要工具。 尽管取得了这些进展，但整合不同类型的数据仍然是一个重大挑战，限制了我们充分利用后基因组学数据财富进行知识和发现的能力。 该项目解决了这一挑战，代表了系统生物学的一个大胆的新方向，可以推广到不同的生物系统建模。 该项目的一个更广泛的影响是软件开发，其目的是通过将可访问的工具交给生物学家来弥合计算生物学和实验生物学之间的差距。

This project develops a comprehensive Bayesian methodology for mathematical modeling, which integrates across diverse data types, accounts for model uncertainty, and incorporates existing prior knowledge about the biological system. The Bayesian models integrate two well-established modeling paradigms: (1) deterministic mathematical models of cellular metabolism and (2) causal graphical models of phenotypes and genotypes. These modeling paradigms are widely used, but suffer major limitations: deterministic models do not reflect the individual variation in fluxes that result from allelic variation of enzymes, and graphical modeling techniques currently ignore all prior information about the biochemical pathway. This project overcomes these limitations by incorporating diverse forms of data and variables, which span the genome, transcriptome, metabolome, and clinical phenotypes. Applications focus on genome-scale metabolic reconstructions of breast and skin cancer cell metabolism to search out novel molecular targeted therapies, side-effects, and unravel the complex mechanisms underlying the disease. The methodology will be implemented in a publicly available software package, which is designed for users with a minimal background in mathematics and statistics. Understanding how genetics influences metabolic and regulatory function is important for the prevention and treatment of disease. Genetics explains in part why some people are more responsive to drug treatments, while others have no response or may experience deadly side effects. Advances in high-throughput phenotyping technologies have made large-scale measurements of molecular traits possible. Mathematical models are widely used to elucidate networks of molecular traits from high-throughput data, and have become important tools for systems biology. Despite this progress, integrating diverse types of data remains a major challenge that has limited our ability to take full advantage of the wealth of post-genomics data for knowledge and discovery. This project addresses this challenge and represents a bold new direction in systems biology, which can be generalized to model different biological systems. A broader impact of this project is the software development, which aims to bridge the gap between computational and experimental biology by putting accessible tools in the hands of the biologist.