High-throughput evolutionary systems biology for expanded genotype-phenotype mapping

我们在多大程度上可以基于基因型预测定量表型？最近的全基因组关联研究表明，遗传性状往往是高度多基因的，并受到许多小效应多态性的影响。这一假设对进化和遗传学领域产生了深远的影响。也就是说，它表明对数量性状位点（qtl）的搜索可以放在一个更广泛的解释框架中，其中表型是由影响它们的因果突变连接的几个生物过程的累积效应的表现。因此，为了理解突变对表型的影响，我们必须考虑基因和调控序列的变化如何影响功能。获得这种映射功能的统计数据，例如，细胞过程对表型的影响程度以及进化对其的选择，是我研究计划的核心问题。在我的建议中，我描述了解决转录过程中这个问题的短期目标，最终目标是将这一策略扩展到其他分子机制，如蛋白质调节。首先，获得任何高维过程的统计数据需要成千上万的综合测量。下一代测序和高通量液体处理机器人技术现在可以以非常低的成本快速测序许多基因组。然而，对DNA水平突变的了解并不能立即转化为对潜在细胞结构如何被扰乱的了解。不幸的是，高通量技术仍然落后于其他基因组学工具，如转录组学和蛋白质组学。利用我们在自动化和反应小型化方面的专业知识，我们将开发大规模询问蜂窝网络的新技术。其次，我们将新的和现有的统计方法应用于新的基因组学数据，以模拟基因型和表型之间的细胞网络。以前，类似的方法已经被用来推断表型是否有遗传成分，而不考虑性状的分子基础。在这里，我们使用这些统计技术来划分对观察到的表型负责的潜在细胞结构的变化。最后，我们使用这些方法直接在实验室中通过询问细胞的系统生物学来测试关于数量遗传学和进化的假设：1)遗传多样性群体中的个体；2)经过数千代进化的细胞谱系之间；3)在没有自然选择的情况下漂流的个体之间。该项目提供了形成表型的分子过程的全面描述，并将这些过程与表型进化联系起来。我们采用跨学科技术，利用机器人液体处理和下一代测序，以弥合对细胞和群体遗传学统计模型中大型数据集的需求。

To what extent can we predict quantitative phenotypes based on genotype? Recent work in genome-wide association studies has proposed that genetic traits are often highly polygenic and influenced by numerous small-effect polymorphisms. This hypothesis has profound consequences for the fields of evolution and genetics. Namely, it suggests that searches for quantitative-trait loci (QTLs) can be placed into a broader explanatory framework where phenotypes are manifestations of the cumulative effects of several biological processes connected by the causal mutations affecting them. Thus, to understand the effect of mutations on phenotypes, we must consider how changes in genes and regulatory sequences affect function in context. Getting at the statistics of this mapping function, for example the extent to which a cellular process can influence a phenotype and be selected on by evolution, is the central question of my research program. In my proposal, I describe short-term objectives that addresses this question for transcriptional processes with the eventual goal of expanding this strategy to other molecular mechanisms such as protein regulation. First, getting at the statistics of any high-dimensional process requires thousands of comprehensive measurements. Next-generation sequencing and high-throughput liquid handling robotics now allow rapid sequencing of many genomes at very low cost. However, the knowledge of mutations at the DNA level do not immediately translate to knowledge of how the underlying cellular architecture has been perturbed. Unfortunately, high-throughput technologies are still lagging for other genomics tools such as transcriptomics and proteomics. Using our expertise in automation and reaction miniaturization, we will develop new technologies to interrogate cellular networks at scale. Second, we apply novel and existing statistical methods for new genomics data to model the cellular network between genotype and phenotype. Previously, similar approaches have been used to infer whether phenotypes had genetic components regardless of the molecular basis of the trait. Here, we use these statistical techniques to partition the variation in underlying cellular architecture that is responsible for the observed phenotypes. Finally, we use these approaches to test hypotheses about quantitative genetics and evolution directly in the lab by interrogating the systems biology of the cell between 1) individuals in a genetically diverse population, 2) between cell lineages over thousands of generations of evolution, and 3) between individuals drifting under no natural selection. This project provides a comprehensive description of the molecular processes that form phenotypes and connects these to phenotypic evolution. We employ interdisciplinary techniques, leveraging robotic liquid-handling and next-generation sequencing to bridge the demand for large datasets in statistical models of the cell and population genetics.