Systematic characterisation of genetically influenced 'omics' phenotypes and disease modules within biological networks

前言：识别与疾病相关的生物学途径和扰动生物学过程的变化的功能特征是了解疾病病因、预后和预防的关键。最近的研究已经成功地确定了遗传变异和潜在致病基因与各种蛋白质、代谢物和脂质的关联，以及它们对与疾病相关的关键生物学途径的影响[1-6]。此外，越来越多的证据表明，不相关的疾病和特征之间存在遗传重叠，这表明疾病有共同的病因[7-8]。拟议项目的目的是通过结合INTERVAL生物资源中的多组学（蛋白质、代谢物和脂质）表型数据、与多组学表型和疾病的遗传关联以及电子健康记录（EHR）来了解疾病的共同原因。背景和目的：如果不构建一个相互作用的生物网络，从这些高维分析中分析大量数据通常是困难的。高斯图形建模（GGM）允许构建生物（组学）网络，并且自动特征检测算法将能够从该网络中提取疾病模块。这里的疾病模块指的是一组蛋白质/脂质/代谢物和相关的途径，这些途径被算法挑选为与疾病相关(例如。冠心病、II型糖尿病等)。在EHR中使用临床措施对这些疾病模块进行进一步的功能表征和计算跟踪，将导致识别与疾病相关的新基因和途径。这也将提高我们对疾病共同成因的理解。ie。如果基因发生了变化，降低了冠心病和糖尿病的风险，但增加了哮喘的风险。该项目将主要关注1。开发一个交互式网络资源，将允许表型数据的调查和遗传关联，在INTERVAL生物资源中测量的表型。2. 开发一种监督学习方法来识别生物网络中的疾病模块研究疾病模块和潜在的生物学途径，使用电子健康记录（EHR）了解疾病的共同病因方法：拟议的工作将侧重于构建多组学数据（蛋白质，代谢物和脂质）的ggm，其边缘表示模型内其他变量条件下两种表型测量之间的部分相关性。元数据包括来自全基因组关联研究（GWAS）的与多组学表型的遗传关联，来自公共数据库（PhenoScanner， SNiPA等）的与疾病的遗传关联以及生物途径（KEGG）的元数据将被添加到网络中，以允许使用监督学习方法识别分子途径和疾病模块。这将是一个特别有用的自动化工具，用于检测生物网络中的疾病模块，并为孟德尔随机化研究（MR）提供信息，以了解疾病的共同病因。这些受基因影响的疾病模块可以通过电子健康记录（HER）表型(例如：诊断，是否存在多发病和药物反应)。这种不可知的方法将提供对组学网络中扰动对医学现象的影响的见解，并确定疾病共享的关键组学表型和途径，从而为治疗干预提供候选靶点。引用:1。Shin S.-Y。et al。Nat,麝猫。(2014) 2。Long， T.等。Nat,麝猫。(2017) 3。Kettunen等人。Commun Nat。(2016) 4。当然，k等人。Commun Nat。(2017) 5。当然，k等人。Nature （2011）Draisma, h.h.m.等人。Commun Nat。(2015) 7。萨瓦尔出版社。《柳叶刀》。(2012) 8。Ferreira et al . PLoS Genetics （2013）

Introduction: Identification of biological pathways associated with diseases and functional characterisation of changes that perturb biological processes is the key to understanding disease aetiology, prognosis and prevention. Recent studies have been successful in identifying the association of genetic variants and potentially causal genes with various proteins, metabolites and lipids and their influence on key biological pathways that are associated with diseases [1-6]. In addition, there is increasing evidence of genetic overlap between unrelated diseases and traits that point to a shared aetiology of diseases [7-8]. The aim of the proposed project is to understand the shared cause of diseases by combining multi-omics (proteins, metabolites and lipids) phenotype data, genetic association with multi-omics phenotypes and diseases and electronic health records (EHR) within the INTERVAL Bioresource. Background and Aim: Analyses of the vast amount of data from these high dimensional analyses is often difficult without constructing an interactive biological network. Gaussian Graphical Modelling (GGM) allows the construction of a biological (omics) network, and an automated feature detection algorithm will enable the extraction of disease modules from this network. Here disease module refers to a set of proteins/lipids/metabolites and associated pathways that are picked by the algorithm as associated with diseases (eg. CHD, Type II Diabetes etc). Further functional characterisation and computational follow-up of these disease modules using clinical measures in EHR will lead to the identification of novel genes and pathways associated with diseases. This will also improve our understanding of the shared cause of diseases. ie. if there's a change in the gene that's associated with reduced risk of CHD and T2D but an increase in Asthma. The project will primarily focus on 1. Developing an interactive web resource that will allow the investigation of phenotype data and genetic association with the phenotypes measured within the INTERVAL bioresource. 2. Developing a supervised learning method to identify disease modules within the biological network 3. Investigate disease modules and underlying biological pathways using electronic health records (EHR) to understand shared aetiology of diseases Methods: The proposed work will focus on constructing GGMs for the multi-omics data (proteins, metabolites and lipids) with edges representing the partial correlation between two phenotype measures conditioned for other variables within the model. Meta-data which include genetic associations with multi-omics phenotypes from Genome-Wide Association Studies (GWAS), genetic association with diseases from public databases (PhenoScanner, SNiPA etc.) and biological pathways (KEGG) will be added to the network to allow the identification of molecular pathways and disease modules using supervised learning methods. This will be particularly useful as an automated tool to detect disease modules within a biological network and inform mendelian randomisation studies (MR) to understand the shared aetiology of diseases. These genetically influenced disease modules can be tested for association across Electronic Health Record (HER) phenotypes (eg. Diagnosis, presence or absence of multi-morbidity and drug responses). This agnostic approach will provide insights into the influence of perturbations within the omics network on medical phenome and identify key omics phenotypes and pathways that are shared by diseases thereby providing candidate targets for therapeutic intervention.References: 1. Shin, S.-Y. et al. Nat. Genet. (2014) 2. Long, T. et al. Nat. Genet. (2017) 3. Kettunen, J. et al. Nat. Commun. (2016) 4. Suhre, K. et al. Nat. Commun. (2017) 5. Suhre, K. et al. Nature (2011) 6. Draisma, H. H. M. et al. Nat. Commun. (2015) 7. Sarwar et.al. Lancet. (2012) 8. Ferreira et.al PLoS Genetics (2013)