Enhanced Gene Identification in Complex Traits Using Kernel Machines

描述（由申请人提供）： 全基因组关联研究（GWAS）已经绘制了数千个常见的性状影响变异，但绝大多数性状位点尚未被发现。该提案的目标是开发和应用超越标准GWAS范式的统计方法，以绘制人类基因组内其他影响性状的变异。我们提出的大多数工具都是基于一个灵活的高维框架，称为内核机器回归，我们已经成功地在GWAS和下一代测序（NGS）研究中成功地应用于复杂性状的强大基因定位。我们相信内核框架固有的灵活性使其成为探索复杂人类特征基因定位新范式的理想选择。目的1提出了新的内核方法，用于整合分析单核苷酸变异数据（来自GWAS和/或NGS）和基因组数据（如基因表达和甲基化模式），我们相信这将为性状定位提供更好的能力。目的2提出了新的核方法，大规模的基因-基因相互作用的分析，在整个基因组，以及一个计算方法，使有效的调整多个测试时，应用这种详尽的测试程序。目的3建立新的核方法用于X染色体上SNV的关联定位。内核机器的灵活性使其成为对该染色体的潜在性别特异性效应建模的理想方法，并且该方法还可以适应随机X失活。目的4提出了一种新的核方法，用于家族内罕见性状影响变异的稳健分析;这种基于家族的设计在当前的罕见变异程序中通常不被考虑。我们将在我们积极参与的大规模数据集上评估这些方法，并将在用户友好的软件中实现这些方法以供公开分发（目标5）。

DESCRIPTION (provided by applicant): Project Summary/Abstract Genome-wide association studies (GWAS) have mapped thousands of common trait-influencing variants yet the overwhelming majority of trait loci have yet to be discovered. The goal of this proposal is to develop and apply statistical approaches that move beyond the standard GWAS paradigm to map additional trait-influencing variation within the human genome. Most of our proposed tools are based on a flexible high-dimensional framework called kernel machine regression, which we have had past success employing for powerful gene mapping of complex traits in GWAS and next-generation sequencing (NGS) studies. We believe the inherent flexibility of the kernel framework makes it ideal for exploring new paradigms in gene mapping of complex human traits. Aim 1 proposes novel kernel methods for integrated analysis of both single-nucleotide variation data (derived from GWAS and/or NGS) and genomic data (such as gene-expression and methylation patterns) that we believe will provide improved power for trait mapping. Aim 2 proposes novel kernel methods for large scale gene-gene interaction analysis across the genome, as well as a computational approach that enables efficient adjustment for multiple testing when applying such exhaustive testing procedures. Aim 3 establishes novel kernel methods for association mapping of SNVs on the X chromosome. The flexible nature of kernel machines makes it ideal for modeling potential sex-specific effects on this chromosome and the methods further can accommodate random X inactivation. Aim 4 proposes novel kernel approach for robust analysis of rare trait-influencing variation within families; such family-based designs are generally not considered in current rare-variant procedures. We will evaluate these methods on large-scale datasets that we are actively involved in and will implement the methods in user-friendly software for public distribution (Aim 5).