Statistical Modeling of Complex Traits in Genetic Reference Super-Populations

描述（由申请人提供）： 模式生物中的遗传杂交在理解医学相关表型的遗传结构中起着至关重要的作用。传统上，这样的杂交往往是小规模的，要么检测遗传效应的能力有限，要么定位的分辨率有限。 因果变量然而，最近，大规模跨学科研究的出现，更便宜的基因分型和人类遗传学的平行进展，刺激了更复杂和强大的实验设计的发展。遗传资源群体（GRP）利用规模经济为遗传研究提供具有成本效益和可复制的平台。该项目涉及迄今为止小鼠遗传学中最大，最雄心勃勃的GRP，协作杂交（CC），以及一系列与之相关或衍生的杂交和设计：多样性远交（DO）杂交，CC重组近交杂交（CC-RIX）和双杂交。在每个单独的交叉上的实验提供了关于目标复杂疾病的遗传结构的不同信息。结合起来，这种遗传参考超级群体（GRSP）可能为小鼠遗传学的交叉研究复制和整合提供了无与伦比的基础。该项目旨在开发统计方法，分别推进这些人群的复杂性状分析的现状，并通过利用连接它们的独特结构，提出开发一个允许它们联合使用的统计框架。 目的1建立一个基于单倍型的数量性状基因座（QTL）分析的贝叶斯概率框架。目的1a开发一个统计软件模块，用于灵活的基于单体型的分析，研究人员可以扩展该模块以模拟丰富多样的设计和疾病类型。目标1b将采用机器学习技术来提供QTL等位基因系列的后验推断。目标1c将纳入多基因效应的贝叶斯建模。 目标2和目标3涉及在目标1的基础上进行联合分析。目的2：开发方法来优化一个人群的随访研究的实验设计，给出另一个人群的结果。目标2a使用该函数为CC/CC-RIX/DO实验的设计提供信息。目标2b使用CC/CC-RIX/DO的部分数据来指导额外数据的收集。目标3探索了在GRSP中联合分析多个群体的模型，使用互补数据集来稳定单个QTL（目标3a）和多个QTL（目标3b）的分析。 这些目标旨在解决多亲本遗传数据，特别是CC、DO、CC-RIX和双杂交种的成本效益设计和有效分析方面的具体和持续挑战。该项目将产生对广泛的模式生物杂交有用的工具，并可应用于任何复杂疾病的遗传研究。

DESCRIPTION (provided by applicant): Genetic crosses in model organisms play an essential role in understanding the heritable architecture of medically relevant phenotypes. Traditionally, such crosses have tended to be on a small scale with either limited power to detect genetic effects or limited resolution to localize causal variants. Recently, however, the emergence of larger-scale interdisciplinary research, cheaper genotyping and parallel advances in human genetics, have spurred the development of more sophisticated and powerful experimental designs. Genetic Resource Populations (GRPs) use economies of scale to provide cost-effective and replicable platforms for genetic studies. This project concerns the largest, most ambitious GRP in mouse genetics to date, the Collaborative Cross (CC), and a series of crosses and designs related to or derived from it: the Diversity Outbred (DO) cross, the CC Recombinant Inbred Cross (CC-RIX) and the diallel. Experiments on each separate cross provide distinct information about the heritable architecture of a target complex disease. In combination, this Genetic Reference Super-Population (GRSP) potentially provides an unparalleled basis for cross-study replication and integration in mouse genetics. This project aims to develop statistical methods that advance the current state of complex trait analysis of these populations separately, and, by exploiting the unique structure that connects them, proposes to develop a statistical framework that allows for their joint use. Aim 1 develops a Bayesian probabilistic framework for haplotype-based analysis of quantitative trait loci (QTL). Aim 1a develops a statistical software module for flexible haplotype-based analysis, which can be ex- tended by the researcher to model a rich variety of designs and disease types. Aim 1b will adapt machine learning techniques to provide posterior inference of the allelic series of a QTL. Aim 1c will incorporate Bayesian modeling of polygenic effects. Aim 2 and 3 concern joint analysis, building on the foundation set by Aim 1. Aim 2 develops methods to optimize experimental design of follow-up studies in one population given results from another. Aim 2a uses the diallel to inform design of CC/CC-RIX/DO experiments. Aim 2b uses partial data on CC/CC-RIX/DO to guide collection of additional data. Aim 3 explores models for jointly analyzing multiple populations in the GRSP, using complementary datasets to stabilize analysis at single QTL (Aim 3a) and across multiple QTL (Aim 3b). These aims address specific and persistent challenges in the cost-effective design and efficient analysis of multiparent genetic data, in particular the CC, DO, CC-RIX and diallel. The project will generate tools useful for a wide range of model organism crosses and can be applied to the genetic study of any complex disease.