Computational Strategies for Quantitative Mapping of Genetic Interaction Networks

描述（由申请人提供）：最近的研究表明，许多疾病，特别是那些普遍困扰我们的人群的疾病，是由多个等位基因之间的相互作用造成的。为了理解这些复杂的表型，最近在模式生物中的实验努力集中在通过工程组合遗传扰动来测量这种相互作用。由于可能的突变体空间巨大，暴力实验研究根本不可行的，因此，迫切需要对遗传相互作用网络进行智能探索的计算策略。本应用程序的具体目标是开发一个计算框架，以利用现有的基因组或蛋白质组学数据来实现组合扰动研究的智能方向。提出的研究的基本原理是，尽管目前对遗传相互作用的了解很少，但现有基因组学和蛋白质组学数据的整合可以使网络模型推断出高通量相互作用筛选的有希望的候选者。使用这样的计算指导应该能够更有效地表征网络结构，并最终更好地理解基因如何促成复杂的表型。基于初步研究的有力发现，这一目标将通过两个具体目标来实现：(1)为基于群体阵列的相互作用分析开发关键的归一化方法和定量模型，以及(2)基于迭代模型改进和最佳相互作用筛选的新型机器学习方法。拟议的研究具有创新性，因为它代表了将基因组数据集成和网络推理技术与大规模实验工作相结合的首次努力之一，其中几个月的实验研究完全基于计算方向。这种方法将使我们深入了解组合扰动如何用于表征全局模块化和组织，更一般地说，将作为其他基因组背景下混合计算-实验策略的原型。

DESCRIPTION (provided by applicant): Recent studies suggest that many diseases, particularly those that commonly afflict our population, result from interactions among multiple alleles. In an attempt to understand these complex phenotypes, recent experimental efforts in model organisms have focused on measuring such interactions by engineering combinatorial genetic perturbations. Due to the enormous space of possible mutants, brute-force experimental investigation is simply not feasible, and thus, there is a critical need for computational strategies for intelligent exploration of genetic interaction networks. The specific objective of this application is to develop a computational framework for leveraging the existing genomic or proteomic data to enable intelligent direction of combinatorial perturbation studies. The rationale for the proposed research is that although current knowledge of genetic interactions is sparse, the integration of existing genomic and proteomic data can enable the inference of network models that suggest promising candidates for high-throughput interaction screens. Using such computational guidance should enable more efficient characterization of network structure, and ultimately, better understanding of how genes contribute to complex phenotypes. Based on strong findings in preliminary studies, this objective will be accomplished through two specific aims: (1) development of critical normalization methods and quantitative models for colony array-based interaction assays, and (2) novel machine learning-based approaches for iterative model refinement and optimal interaction screen selection. The proposed research is innovative because it would represent one of the first efforts to couple genomic data integration and network inference technology with a large-scale experimental effort, where several months of experimental investigation are based entirely on computational direction. Such an approach will yield insight into how combinatorial perturbations can be used to characterize global modularity and organization, and more generally, would serve as a prototype for hybrid computational-experimental strategies in other genomic contexts. PUBLIC HEALTH RELEVANCE: Many common diseases result from interactions among multiple genes. One approach to studying multigenic interactions is to introduce combinations of mutations in model organisms and observe how they affect the cell. This project proposes to develop computational strategies to guide and interpret these combinatorial perturbation studies, which will ultimately help us better understand and treat multigenic diseases.