Solar-Eclipse Computational Tools for Imaging Genetics

 描述（由申请人提供）：本申请将为新兴的成像遗传学领域提供急需的分析方法。我们的重点是SOLAR-Eclipse集成资源套件的第二阶段开发，用于遗传和表观遗传分析，如遗传力，多效性，数量性状位点连锁（QTL-L），全基因组关联（GWA）和全基因组测序（WGS），基因表达和甲基化分析，这些分析针对来自结构和功能神经成像数据的性状进行了优化。在第一个短暂而密集的资助期间（2.5年），我们展示了SOLAR-Eclipse在成像遗传学应用中的实用性，并与三个主要的NIH脑成像计划建立了强大的“拉/推”合作：NIH大数据2知识（BD 2K）通过荟萃分析增强神经成像遗传学（ENIGMA），人类连接组项目（HCP）和中风遗传学网络（SiGN）。在第一个资助期内，我们发布了12个主要的软件更新，并撰写和合著了47篇手稿。我们在成像遗传学会议（2012年，2013年，2014年，2015年）和人脑映射组织会议（2012年，2013年，2014年）的遗传成像研讨会上建立了关于使用SOLAR-Eclipse的年度研讨会。我们在大数据合作研究的“拉/推”精神中构建了这一更新，其中“拉”指的是我们团队开发的新工具，“推”指的是与大数据合作伙伴合作应用和测试尖端分析。我们建议将SE开发的下一阶段重点放在我们的大数据合作伙伴确定的需求上。我们提出了三个“拉动”目标（1-3），以开发领先的和使成像遗传学分析技术，高性能计算量身定制的独特成像遗传学的挑战，和新的数据格式。在“推动”AIM 4中，SOLAR-Eclipse团队与成像遗传学合作，ENIGMA，HCP，SiGN，IMAGEN等合作，通过合作研究“推动”科学的发展。AIM 1以高性能计算为中心，旨在实现对基于家族的样本（如HCP）中的体素成像特征的实时GWA/WGS分析。通过将用于快速近似似然分析的新型数据转换与图形处理器单元（GPU）计算相结合，与传统的最大似然计算方法相比，我们将实现约105-6倍的计算加速。高性能计算将需要一种新的数据格式来存储成像遗传学数据。在AIM 2中，我们建议起草Gen.Gii数据格式和应用程序编程接口（API），为成像遗传学分析进行优化，并记录成像遗传学数据分析工作流程的起源。基于成像遗传学专家工作组创建的初稿，我们将寻求广泛的社区投入，以确保Gen.Gii标准将被接受。在AIM 3中，我们建议将新开发的经验亲缘关系技术方差分量核用于成像遗传学应用。经验亲缘关系方法直接从全基因组数据计算个体间的“遗传距离”，并根据“经验亲缘关系”划分性状方差，如计算精神分裂症患者白色物质完整性的加性遗传方差，该方差是由基因组调控区贡献的。经验亲属关系方法将被推广到进行经典的遗传方差分析的成像表型，包括他们的遗传力，多效性，罕见的变异和数量性状连锁分析在“无关”（但实际上是远亲）的主题（2）。在AIM 4中，我们将进行合作研究， 我们的大数据合作伙伴ENIGMA、HCP和SiGN收集了大量不同的样本。这种新方法的合作试点和磨练将有助于推广和传播我们为个人成像遗传学实验室开发的技术。

 DESCRIPTION (provided by applicant): This application will provide urgently needed analytical methods to the emerging field of imaging genetics. Our focus is on phase 2 development of SOLAR-Eclipse integrated suite of resources for genetic and epigenetic analyses such as heritability, pleiotropy, quantitative trait loci-linkage (QTL-L), genome-wide association (GWA) and Whole-Genome Sequencing (WGS), gene expression, and methylation analyses optimized for traits derived from structural and functional neuroimaging data. During the first short and intensive funding period (2.5 years), we demonstrated the utility of SOLAR-Eclipse for imaging genetics applications and developed strong "Pull/Push" collaboration with three major NIH brain imaging initiatives: the NIH Big Data 2 Knowledge (BD2K) Enhancing NeuroImaging Genetics through Meta-Analysis (ENIGMA), Human Connectome Project (HCP) and Stroke Genetics Network (SiGN). During the first funding period, we released 12 major software updates and authored and co-authored 47 manuscripts. We established an annual workshop on the use of SOLAR-Eclipse at the Imaging Genetics Conference (2012, 2013, 2014, 2015) and at the genetic imaging workshop at the Organization for Human Brain Mapping conference (2012, 2013, 2014). We structure this renewal in the "Pull/Push" sprit of collaborative Big Data research, where "Pull" refers to development of novel tools by our team and "Push" refers to collaboration with Big Data partners to apply and test cutting edge analyses. We propose to focus the next phase of SE development at the needs identified by our Big Data partners. We propose three "Pull" AIMS (1-3) to develop leading and enabling imaging genetics analysis techniques, high performance computing tailored to unique imaging genetics challenges, and novel data formats. In "Push" AIM 4, SOLAR-Eclipse team partnered with imaging genetics collaborations, ENIGMA, HCP, SiGN, IMAGEN and others to "Push" the state of science through collaborative studies. AIM 1 centers on high performance computing with the aim of achieving real-time GWA/WGS analyses of voxel-wise imaging traits in family based samples such as HCP. By combining novel data transformations for fast approximation of likelihood analyses and Graphics Processor Unit (GPU) computing, we will achieve ~105-6-fold computation acceleration as compared with traditional, maximum likelihood calculation methods. High performance computing will require a new data format for storage of imaging genetics data. In AIM 2, we propose to draft Gen.Gii data format and application programming interface (APIs) optimized for imaging genetic analyses, as well as recording the provenance of imaging genetics data analysis workflows. Building off a first draft created by a working group of imaging genetics experts, we will seek broad community input to ensure that Gen.Gii standard will be embraced. In AIM 3, we propose to integrate newly developed empirical kinship techniques variance component kernel use in imaging genetics applications. Empirical kinship methods calculate "genetic distances" among subjects directly from genome-wide data and partition the trait variance based on the "empirical kinship"; for example, computing additive genetic variance in white matter integrity in schizophrenia patients that is contributed by the regulatory regions of genome. Empirical kinship methods will be generalized to perform classical genetic variance analyses of the imaging phenotypes, including their heritability, pleiotropy, rare variant and quantitative trait linkage analyses in "unrelated" (but actually distantly related) subjects (2). In AIM 4, we will execute collaborative studies to fine tune novel methods in large and diverse samples assembled by our Big Data partners: ENIGMA, HCP and SiGN. This collaborative piloting and honing of novel methods will serve to popularize and disseminate our developments for individual imaging genetics labs.