High-performance mixed model toolset for integrative omics analysis of big data

用于大数据综合组学分析的高性能混合模型工具集

• 批准号: 9312511
• 负责人: JEFFREY R O'CONNELL
• 金额: $ 58.48万
• 依托单位: UNIVERSITY OF MARYLAND BALTIMORE
• 依托单位国家: 美国
• 财政年份: 2017
• 资助国家: 美国
• 起止时间: 2017-04-15 至 2020-03-31
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/9312511
• 关键词: Algorithms; Attention; Beds; Big Data; Biological; Biological Models; Biology; Cloud Computing; Code; Collaborations; Communities; Complex; Complex Genetic Trait; Computer software; Data; Data Analyses; Data Set; Development; Disease; Epigenetic Process; Equation; Family; Genetic; Genetic Epistasis; Genomics; Genotype; Goals; Health; Heterogeneity; Human; Investments; Memory; Meta-Analysis; Methodology; Mitochondria; Modeling; National Heart, Lung, and Blood Institute; National Human Genome Research Institute; Performance; Phase; Phenotype; Play; Population; Process; Publishing; Research; Research Personnel; Resource Allocation; Resources; Role; Sample Size; Sampling; Sequence Analysis; Shapes; System; Systems Biology; Technology; Testing; Time; Trans-Omics for Precision Medicine; Variant; Weight; Work; analytical method; animal breeding; base; biological systems; cloud based; cohort; cost; data access; data management; data space; epigenome-wide association studies; file format; flexibility; genetic analysis; genetic pedigree; genomic data; improved; insight; large scale production; method development; novel; novel strategies; precision medicine; pressure; rare variant; response; scale up; simulation; simulation software; terabyte; tool; trait; virtual; whole genome; working group

PROJECT SUMMARY/ABSTRACT The recent large scale production of whole genome sequence and other multi-omics in TOPMed and other projects calls for parallel development of comprehensive, powerful and flexible toolset capable of large data management, analysis and integration. Mega/integrated analyses are essential to fully utilize these data to elucidate the complexity of the biological mechanisms and advance our understanding of complex trait biology to drive precision medicine. TOPMed estimates that the VCF for 60,000 subjects will contain 400M variants and require 100TB of space, and much of our current genetic analysis toolset does not scale up to these data sizes. For rare variant analysis, mixed model mega analysis is more powerful than meta-analysis as mega analysis can include additional random effects to account for genetic relatedness between all subjects and cross-study phenotypic, genetic and environmental heterogeneity. However cross-study mega analysis within the mixed model is still an uncharted territory. We believe mega analysis will spur more creative analysis approaches provided the needed toolsets are available. In cloud computing “time is money”, and new approaches are required to solve structural differences in resource allocation and data access compared to local computing. MMAP (Mixed Models for Analysis of Pedigrees/Populations) is robust mixed model software that already published mixed model analysis on a sample size of 90,000 that included dominance variance and developed a cloud-efficient version of mixed model rare variant analysis. The goal of this proposal is to further expand and improve this toolset to deliver to the research community a flexible, versatile, and comprehensive cross-platform mixed model toolset scalable to efficient local and cloud analysis of large WGS and omics data. We plan to implement several new features in our toolset including: 1) Efficient binary genotype file format for optimal storage of terabyte VCF genotypes. 2) Large-scale modeling of non-additive variation such as dominance, X- lined, mitochondrial and epistasis. 3) Optimized rare variant analysis with flexible integration of annotation and variant weighting resources. 4) Optimized expression/epigenome-wide association (EWA) analysis. 5) Comprehensive multi-omics integration into the mixed model as fixed and random effects. 6) Development of a multi-omics simulation software to guide systems biology modeling. 7) Integrating mixed model equations for prediction from animal breeding. This proposal will deliver the research community an analysis toolset that will push research boundaries well beyond additive SNP association to a space filled with complex biological fixed and random effects models integrating the full spectrum of multi-omics data. We plan to develop a multi-omics simulation tool to better understand the complex evolutionary processes that shape the complex trait landscape. Our toolset will be extensively shaped by collaboration with TOPMed working groups to meet analysis priorities and develop analysis plans. Our toolset will surely evolve in novel and unexpected directions in response to new ideas and challenges as we dive deeper into this unique data set.

项目总结/摘要 最近大规模生产的全基因组序列和其他多组学在TOPMed和其他 项目要求并行开发能够处理大数据的全面、强大和灵活的工具集 管理、分析和整合。大型/综合分析对于充分利用这些数据至关重要， 阐明生物机制的复杂性，并推进我们对复杂性状生物学的理解， 推动精准医疗TOPMed估计，6万名受试者的VCF将包含4亿个变体， 需要100 TB的空间，而我们目前的大部分遗传分析工具集都无法扩展到这些数据大小。 对于罕见变异分析，混合模型的大分析比荟萃分析更强大，因为大分析可以 包括额外随机效应以解释所有受试者和交叉研究之间的遗传相关性 表型、遗传和环境异质性。然而，混合内部的交叉研究大型分析 模型仍然是一个未知的领域。我们相信大分析将激发更多创造性的分析方法 前提是所需的工具集可用。在云计算中，“时间就是金钱”，新的方法 与本地计算相比，需要解决资源分配和数据访问的结构差异。 MMAP（用于谱系/群体分析的混合模型）是一个强大的混合模型软件， 发表了对90，000个样本量的混合模型分析，其中包括显性方差，并开发了一个 混合模型罕见变异分析的云高效版本。该提案的目标是进一步扩大和 改进此工具集，为研究社区提供灵活、通用和全面的跨平台 可扩展的混合模型工具集，可对大型WGS和组学数据进行高效的本地和云分析。我们计划 在我们的工具集中实现了几个新功能，包括：1）高效的二进制基因型文件格式， 存储TB级VCF基因型。2)非加性变异的大规模建模，如显性、X- 线粒体和上位性。3)优化的罕见变异分析，灵活集成注释和 不同的权重资源。4)优化表达/表观基因组全关联（EWA）分析。第五章） 综合多组学整合到混合模型中作为固定和随机效应。6)发展 多组学模拟软件指导系统生物学建模。7)混合模型方程的积分 动物育种的预测。该提案将为研究界提供一个分析工具集， 将研究边界远远超出添加剂SNP关联，进入充满复杂生物固定的空间。 和随机效应模型整合了多组学数据的全谱。我们计划开发一个多组学 模拟工具，以更好地了解复杂的进化过程，塑造复杂的性状景观。 我们的工具集将通过与TOPMed工作组的合作进行广泛的塑造，以满足分析优先级 制定分析计划。我们的工具集肯定会朝着新的和意想不到的方向发展， 想法和挑战，因为我们深入研究这个独特的数据集。