Computation and functional significance of multi-phenotype genetic interaction ma

描述（由申请人提供）：两个遗传位点之间的上位性表明它们之间存在相互作用，即对表型的综合影响与基于其个体影响的预期相反。计算机模拟和高通量技术的可用性使得同时探索几个上位交互成为可能，从而产生上位交互网络。这些网络在解释通路功能和进化适应，以及在多性状遗传疾病的研究和药物联合治疗的发展中发挥着越来越重要的作用。由于这些原因，越来越多的实验和计算工作集中在上位交互数据的收集、模拟和分析上。然而，一个经常被忽视的问题是选择表型的重要性，相对于两个基因之间的相互作用是确定的。对单一表型的限制很大程度上是探索许多可能的遗传变异和表型的组合复杂性的结果。在这里，我们建议利用实验驱动的酵母酵母代谢网络的硅基因组尺度模型来生成和研究生物网络中所有可能的表型和扰动的第一个上位相互作用图。对系统的扰动将是代谢酶基因的缺失，表型将由系统的所有可计算变量组成，即所有细胞内和运输代谢反应速率（通量）。具体而言，我们将在一组预定义的环境条件下计算所有单次和双次扰动（基因缺失）的所有通量（表型），选择适当的上位性度量，然后推导相互作用的三维矩阵（目标1）。所有通量表型的集合将构成包含代谢基因之间依赖关系的功能指纹，可用于规划后续实验和生物医学相关应用（如预测疾病和开发疗法）。接下来，我们将通过使用高通量方法构建合适的菌株和一套强大的检测方法来测量大量单和双酵母突变体中选择的通量表型，从而测试大量这些预测（目标2）。最后，我们将实现一个用于多表型上位分析的在线平台，通过该平台，用户不仅可以下载数据和软件，还可以执行新的计算并生成用户特定的预测和地图（目标3）。我们期望，与单一表型图相比，我们的多表型图将揭示新的相互作用，并将传达过程之间关系的更丰富的观点。我们所提出的工作将为生物系统中多表型上位相互作用数据的分析奠定理论、计算和交互式可视化基础。

DESCRIPTION (provided by applicant): Epistasis between two genetic loci indicates an interaction between them, i.e. a combined effect on phenotype that defies expectations based on their individual effects. The availability of computer simulations and high-throughput technologies makes it possible to explore simultaneously several epistatic interactions, giving rise to epistatic interaction networks. These networks play an increasingly central role in explaining pathway functions and evolutionary adaptation, as well as in the study of multi- trait genetic diseases and in the development of drug combination therapies. For these reasons, a growing number of experimental and computational efforts focus on the collection, simulation and analysis of epistatic interaction data. Yet, an often neglected matter is the importance of the choice of the phenotype relative to which the interaction between two genes is defined. The limitation to a single phenotype is largely a consequence of the combinatorial complexity of exploring many possible genetic variants and phenotypes. Here, we propose to take advantage of experimentally-driven in silico genome- scale models of the metabolic network of the yeast S. cerevisiae to generate and study the first epistatic interaction map for all possible phenotypes and perturbations in a biological network. The perturbations to the system will be the deletions of metabolic enzyme genes, and the phenotypes will consist of all computable variables of the system, i.e. all intracellular and transport metabolic reaction rates (fluxes). Specifically, we will compute all fluxes (phenotypes) for all single and double perturbations (gene deletions) under a set of predefined environmental conditions, choosing an appropriate epistasis metric, and then deriving the three-dimensional matrix of interactions (Aim 1). The set of all flux phenotypes will constitute a functional fingerprint containing dependencies between metabolic genes, which can be used for planning subsequent experiments and for biomedically relevant applications (like predicting disease and developing therapies). Next, we will test a significant number of these predictions by using high throughput methods to construct the appropriate strains and a robust set of assays to measure selected flux phenotypes in a large number of single and double yeast mutants (Aim 2). Finally, we will implement an online platform for multi-phenotype epistasis analyses through which users will be able not only to download data and software, but also to perform novel calculations and generate user-specific predictions and maps (Aim 3). We expect that, compared to single phenotype maps, our multi-phenotype map will reveal novel interactions and will convey a much richer view of the relationships between processes. The work we are proposing will lay the theoretical, computational and interactive visualization foundations for the analysis of multi-phenotype epistatic interaction data in biological systems. PUBLIC HEALTH RELEVANCE: Complex networks of interactions between genes are ubiquitous in biological systems, posing fundamental barriers that severely limit our capacity to address major biomedical challenges, such as complex genetic diseases as well as drug interactions and side- effects. This proposal will address this problem by generating a new computational representation of genetic networks, which will help predict, visualize and experimentally screen biomedically relevant interactions.