Strategy to innovate the Korea Biobank to effectively respond to infectious diseases

김희수 Graduate School, Yonsei University 2024 국내박사

한국인체자원은행에서의 임상문서표준구조 등록 시스템 구현

예정훈 경북대학교 대학원 2009 국내석사

The Biobank has been developed to allow that authorized researchers can search and access CDA(Clinical Document Architecture) documents for the clinical information associated with particular human biomedical resources of their interests. It needs both of definition and management system of metadata of CDA document related to human biomedical resource. In addition, it should support interoperability among biobanks. In this paper we propose a registry system for the biobank based on the above requirements. It manages and stores extended metadata of registered CDA documents, which extends IHE(Integrating the Healthcare Enterprise) XDS(Cross Enterprise Document Sharing) Document metadata. For interoperability among biobanks, it uses ebXML(Electronic Business using eXtensible Markup Language) as a base framework. In the implementation, we adopted ORM(Object-relational mapping) tool for database-independence, Java for platform-independence, and RESTful (Representational State Transfer) web services for accessibility from heterogeneous operating system. With this registry system managing extended metadata of CDA documents, we can use CDA documents efficiently. In performance evaluation, it is shown that this system is faster in registration time, but slower in query response time than ebxmlrr, open-source registry based on SOAP(Simple Object Access Protocol); therefore, performance in querying should be improved.

Prediction of Disease Risk in a Large-Scale Biobank with Graph Neural Networks

김도균 서울대학교 대학원 2023 국내석사

The availability of advanced technology, such as machine learning and large-scale data, has facilitated significant improvements in disease research. This research focuses on the UK Biobank, a large-scale biobank in the UK, which provided metabolomics data for a substantial number of participants. The integration of this data, along with clinical records and genetic information, was used to predict 22 different diseases. We propose a disease prediction model based on a Graph Neural Network that utilizes the interrelations of metabolites. Since direct interaction information between metabolites was not available, a network was created using genetic correlations obtained from genome-wide association study data. The metabolomics graph model follows a multitask learning approach, with a shared GNN network and disease-specific head networks. The model’s performance is evaluated against a LASSO regression baseline model, with the metabolomics graph model demonstrating improved prediction performance in certain diseases. 기계 학습과 대규모 데이터와 같은 진보된 기술의 활용 가능성으로 인해 질병 연구에서 상당한 발전이 이루어졌습니다. 본 연구는 영국의 대규모 바이오뱅크인 UK Biobank에서 제공하는 다수 참여자의 대사체 데이터에 초점을 두고 있습니다. 이 데이터로 임상 기록 및 유전 정보를 통합하여 22가지 다른 질병을 예측을 하였습니다. 본 연구에서는 유전 연관 분석(Genome-wide Association Study) 데이터를 통해 유전적 상관성을 계산하였습니다. 직접적인 대사체간 상호작용 정보가 없는 상황에서 계산한 유전적 상관성을 이용해 네트워크를 생성하였고 그래프 신경망을 기반으로 한 질병 예측 모델을 제안합니다. 대사체 그래프 모델은 공유 그래프 신경망과 개별 질병 신경망으로 구성되며 멀티태스크 학습 방식을 따릅니다. LASSO 회귀 모델과 비교하여 모델의 성능을 평가하였으며 대사체 그래프 모델은 일부 질병에서 예측 성능이 향상되었음을 보여주었습니다.

Leveraging Computational Methods and Electronic Health Records-Linked Biobank Data in Oral and Craniofacial Health Research

Venkateswaran, Vidhya University of California, Los Angeles ProQuest Dis 2023 해외박사(DDOD)

소속기관이 구독 중이 아닌 경우 오후 4시부터 익일 오전 9시까지 원문보기가 가능합니다.

Bioinformatics and computational methods play an important role in advancing medical research with their ability to leverage large datasets, including data from electronic health records (EHR) linked biobanks. Precision medicine can benefit from leveraging a more comprehensive picture of a patient's genotypes and phenotypic presentation for targeted interventions and treatment planning. In this work, I discuss the applications of bioinformatics methods in the UCLA ATLAS biobank, in evaluating craniofacial traits and their risk factors: specifically, head and neck cancer and tobacco use disorder. First, I describe phenome-wide and lab-wide association analysis pipelines that leverage the breadth of the available information in the biobank, and the results of preliminary investigations of the phenome-wide and laboratory-wide associations of a genetic predisposition to tobacco use disorder. Next, I present the results of an evaluation of the predictive performance of a tobacco use polygenic score across different genetic ancestry groups and further discuss the differences in disease presentations in tobacco use-predisposed individuals with and without a history of the associated tobacco use behavior. Next, I employ these pipelines and statistical methods in the examination of the interplay of serum bilirubin, tobacco use, head and neck, and lung cancer. I present the results of this project, examining the effect of environmental and genetic factors on serum bilirubin and associations with head and neck cancer and lung cancer. Lastly, I propose a research project to examine the germline risk factors for oropharyngeal cancer and discuss the future directions of this work.

인체유래검체 사용과 관련된 생명윤리 이슈들에 대한 한국병리의사들의 인지도 조사

박영준 고려대학교 대학원 2011 국내박사

Purpose : Since the ‘Korean Bioethics and Safety Act(KBSA)’ has been effective in 2005, a new expanded revision for its systematic management has been expected in the near future. Pathologists are experts dealing with human derived specimen and the central operating body of biobanks. However, there was no study on the recognition of Korean pathologists about bioethical issues and management of biobanks in Korea. This study was performed to assess the knowledge and perceptions of bioethical issues related to the usage of human biological materials (HBM) and KBSA among Korean pathologists. Materials and Methods : Subjects were attendee at the conferences held by Korean society of pathologists and the Korean society for cytopathology in 2009; pathology specialists (n=394) and pathology residents (n=109). The questionnaire was composed of 4 categories including ‘KBSA-related issues‘ (15 items), ‘biobank related issues’ (7 items), ‘debating issues about use of human specimens’ (10 items), and ‘history of bioethics’ (4 items). The Results were compared between pathology specialists and pathology residents. Results : In a self-reporting survey of 503 respondents, it was revealed that the Korean pathologists were more likely to estimate negatively of the ethical consciousness of the Korean doctors or scientists compared to that of the scientists in advanced countries(P < 0.01). As for the most issues, both the specialists and pathologists demonstrated appropriate understandings. Most of the pathology specialists and residents answered that a written consent was essential for the use of HBM. However, the majority of the specialists and some of the residents, too, answered they thought an unlimited use of HBM was possible if it was obtained from a donor who died without bereaved family. Besides, most pathologists, either specialists or resident, considered that biobanks, rather than the donors, have a right to own HBM(P < 0.05). That they showed a rather poor understanding of conflict of interests was another problem. Conclusions: Korean Pathologists showed generally high understanding about bioethical issues related to the usage of HBM and KBSA. But, major factors that explain ethical and legal perceptions should be continuously educated to avoid potential harm for human subjects.

Identification of genetic traits associated with inflammatory bowel disease using the UK Biobank database : For the 2,145 naive IBD patients of United Kingdom

홍진희 서울대학교 대학원 2022 국내석사

Background: The IBD continuously increased across all over the industrialized societies. Estimated prevalence of IBD was from 3.7 million to more than 6.8 million between 1990 and 2017. Northern Europe especially United Kingdom and America show highest incidence of IBD. As a human body being a host of microbes, they attack our body causing an inflammation response and eventually lead to IBD. The cause of IBD is not yet known, but it is presumed to be caused by genetic, immune, and environmental factors. By previous study, pathologically causes of IBD are improper response of a inadequate mucosal immune system, impaired innate immune mechanisms of the epithelia layer, wrongly recognizing antigen, atypical antigen presenting cells, overreactive T cell and environmentally problems like food and stress. But still, it’s lack of explainable in biologically. Needs for studies to understand pathway by discover IBD associated gene function. Genes associated study for IBD was previously studied from 1981. Many genes related to IBD reacting to several pathway such as barrier function, epithelial restitution, innate immune regulation, autophagy. Well known genes, IL23R and NOD2 regulates immune responses. CDH1, GNA12 and PTPN2 react to barrier integrity. CARD9 integrates signals from many innate immune receptors that recognize viral, bacterial, and fungal motifs. IL-23, IL-6 and IL-17 crucial role in Th17 cell proliferation. Objective: The aim of this study is to find out associated gene with IBD by using UK BioBank with white population more than 2,000 patients. Also, categorize subgroups in IBD and compared each group gene function and evaluate lab information with candidate genetic loci genotype level. Methods: In this study, using the UK Biobank genetic and clinical data from 500,000 subjects. Detailed medical data in Hospital Episode Statistical data about diagnosis, General Practice data for drug prescription and Lab data with 67 blood test and biochemistry test results. As a definition of IBD, using ICD code and number of counter visits. For a sensitivity analysis five drugs for treating IBD were used and operation for measuring severity of IBD. After patient selection used their whole exome sequencing data to discover candidate of genetic loci and genes associated with IBD. Two kind of variant level test were performed. As a frequency level, Cochran Armitage test was used for additive and the other models (allelic, recessive, and dominant) were tested by Fisher’s exact test. In a genotype level used multiple logistic regression adjusted with sex and age. For a sensitivity group, the type of IBD-specific medication and severity were additionally adjusted. As a gene level test which aggregates variants in the same gene were scored by Gene Variant wise Burden test and tested by multiple logistic regression. All candidates were passing adjusted P value under 0.05. Results: Study groups were diagnosed with ICD 9 and 10 version 2,145 IBD cases, 714 CD cases and 1,431 UC cases and 573 sensitivity IBD cases group. In total, 13 disease-associated loci (4.33ⅹ10^(−8) < P < 1.05ⅹ10^(−6)) were identified in a IBD group. IL23R(Interleukin 23 Receptor), rs11209026 (P = 4.33ⅹ10^(−8); OR 0.63) driving association in both variant-level and gene-level test(P = 1.16ⅹ10^(−8)). IL23R plays critical role in IBD related to protection against the development of CD and UC. CARD9(Caspase Recruitment Domain Family Member 9), rs4077515 (P = 3.54ⅹ10^(−7); OR 1.20) functions in the immune response against microorganisms. Another essential in pathology of IBD is GPX1 (Glutathione Peroxidase 1), rs1050450 (P = 2.46ⅹ10^(−5); OR 1.39). Roles in reduction of oxidase and peroxide. Lastly CFB (Complement Factor B), rs4151651 (P = 3.54ⅹ10^(−7); OR 1.20) is highly upregulated downstream of TLR3(Toll like receptor 3). TLR3 is one of the IBD pathogenesis role in innate immune system. Functional analysis was performed by DAVID tool. As a result, IBD function was clustered most highly in membrane and signal. Additionally, linear regression was performed as a clinical analysis between IBD patients with CFB (rs4151651) carrier and other clinical traits. IBD patients with CFB carrier was linked with traits as apolipoprotein A, Aspartate aminotransferase, C-reactive protein, HDL cholesterol, potassium in urine, and alanine aminotransferase. CRP levels are used in serum marker of IBD to follow up patient’s inflammation react. In this analysis, when CFB variant has carrier induced highest level of CRP. Conclusion: The availability of big scale of whole exome sequence data makes possible to insight the contribution of nonsynonymous variants affects into IBD. Moreover, biological function and checking discovered genetic loci’s genotype effect on IBD prognostic trait helps to understand effect size of carrier and non-carrier gene. Comparing each CD and UC associated genes’ function helps to understand functional difference of the two group. 연구 배경: 산업화가 진행되면서 염증성 장질환은 전세계적으로 높은 발병률을 가지게 되었다. 살펴본 결과 1990년부터 2017년까지 질병 발생수는 370만명에서 680만명까지 증가였으며 전세계적으론 85.1%의 발병률이 증가하게 되었다. 일반적으로 가장 높은 발병률을 가진 국가는 북유럽의 영국과 북미로 보고되고 있다. 염증성 장질환은 외부 환경에 대해 민감한 숙주의 공생 미생물에 대한 공격에 대해 반응하고 유전적인 영향 등 지속적인 염증 반응으로 위장에서부터 장까지 염증이 발생하는 것으로 알려져 있다. 염증을 일으키는 원인은 여러 가지가 있다. 점막 면역 체계에 결함이 생겨 부적절하게 반응하는 것, 상피층의 선천적 면역 메커니즘의 교란, 잘못된 항원 인식, 비정형 항원 제시 세포, 과민성 T 세포 및 음식, 스트레스와 같은 환경 문제들로 인하여 염증 반응이 일어난다. 그러나 아직까지 염증성 장질환이 일어나는 생리학적인 분명한 원인은 알려지지 않은 것으로 추가적인 연구를 통해 질병의 중요성을 인식해야 된다는 점이 대두되고 있다. 염증성 장질환과 연관성 있는 유전자와 관련된 연구는 1981년에서부터 현재까지 꾸준히 연구되었다. 질병과 관련된 많은 유전자는 장벽 기능, 상피 회복, 선천적 면역 조절, 자가포식과 같은 여러 경로와 연관성이 있었다. 잘 알려진 유전자로는 IL23R과 NOD2로 면역 반응을 조절한다고 알려져 있다. 또한 장벽 무결설과 기능이 있는 CDH1, GNA12 및 PTPN2또한 연관성이 있다고 알려져 있다. CARD9는 바이러스, 박테리아 및 곰팡이 모티프를 인식하는 많은 선천성 면역 수용체의 신호를 통합한다. IL-23, IL-6 및 IL-17은 Th17 세포 증식에 중요한 역할을 한다. 면역 체계와 관련된 유전자들에 변이가 생길 경우 장 기능에 결함이 가 염증성 장질환과 같은 염증 반응을 지속적으로 일어나게 할 수 있다. 연구 목적: 본 연구의 목적은 2,000명 이상의 백인 환자를 대상으로 50만 이상의 환자 정보를 가지고 있는 UK Biobank 데이터 셋을 활용하여 염증성 장질환과 관련된 유전자를 찾고, 후보군들이 genotype heterozygosity에 따라 예후 영향에 서로 다른 영향을 미치는지 확인한다. 또한 염증성 장질환의 하위 그룹 크론병과 궤양성 대장염으로 나눠 각 그룹의 후보군과의 비교와 더불어 생리학적 차이들을 살펴보았다. 방법: 이 연구는 500,000명의 기증자의 유전체 데이터를 포함하는 UKBiobnak 데이터셋을 사용하였다. 피험자들의 병원 진단 기록과 더불어 약물 처방 기록, 일반 진료 데이터 및 67개의 혈액 검사와 생화학 검사 결과가 포함된 실험실 데이터들을 활용하였다. 임의적으로 지정한 실험군에 대한 정의는 ICD 9과 10 진단 코드로 정의 내려진 염증성 장질환과 외래방문 기록 1회 이상인 환자군을 뽑았으며, 크론병과 궤양성 대장염도 또 다른 그룹군으로 나누었다. 염증성 장질환에 대한 또 다른 조작적 정의로는 ICD 코드와 더불어 질환 치료제로 알려진 5가지 약물을 한 번 이상 사용한 사람을 순수한 염증성 장질환군으로 나누었다. 유전적 연관성을 살펴보기 위한 엑솜 시퀀싱 데이터를 사용하였고, 두 종류의 변이 수준의 테스트와 유전자 수준의 테스트를 진행하였다. 변이의 빈도 수준의 테스트는 Additive 모델의 경우 Cochran Armitage test를, 나머지 dominant, recessive, allelic 모델의 경우 Fisher’s exact test를 진행하였다. Genotype 순준의 변이 테스트는 multiple logistic regression 방법으로 나이와 성별을 조정하였다. 유전자 수준 테스트는 Gene level burden test 정의에 따라 변이별 GVB 점수를 매기고 같은 유전자에 속한 변이들을 하나의 유전자의 값으로 통합한 뒤 multiple logistic regression 테스트를 진행하였다. 모든 후보군들은 보정된 p value값이 0.05 미만인 유전자와 변이들을 추출하였다. 결과: 네 군의 연구 코호트는 진단코드로 정의 내려진 2,145명의 염증성 징질환 환자군과, 714명의 크론병, 1,431명의 궤양성 대장염 환자군 및 약제와 진단 코드로 정의된 573명의 염증성 장질환군으로 나눴다. 총 13개의 질병 관련 유전자 좌위(4.33ⅹ10^(-8) < P < 1.05ⅹ10^(-6))가 염증성 장질환과 연관성이 있는 것으로 확인되었으며 유전자 수준, 변이 수준에서 모든 테스트를 통과한 후보군은 IL23R, rs11209026(P = 4.33ⅹ10^(−8); OR 0.63)이었다. IL23R은 크론병과 궤양성 대정염의 발달에 대한 보호와 연결되어있으며, 염증성 장질환에서 중추적인 역할을 합니다. IL23은 STAT3, STAT4에 결합하여 단일화를 감소시키고 T 세포, NK 세포 및 대식세포를 매개한다. CARD9, rs4077515(P = 3.54ⅹ10^(−7); OR 1.20)는 미생물에 대한 면역 반응에서 기능합니다. GPX1, rs1050450 P = 2.46ⅹ10^(−5); OR 1.39)는 염증성 장질환의 병리학에서 또 다른 필수적인 역할을 하는 유전자로 알려져 있다. Oxidase와 peroxide의 환원 역할을 하여 염증 반응을 감소시키며, GPX1과 GPX2 유전자를 knockout시킨 쥐 실험체의 경우 회장과 결장에서 점막 염증의 발병률이 높아진다는 이전 연구가 있다. 그 외 4개의 크론병 관련 변이체가 발견되었다. LIG ,rs3730947(P = 2.92ⅹ10^(−6); OR 15.1)은 ATP 의존성 DNA리가아제의 구성원을 인코딩하고 DNA복제, 재조합 및 염기 절단 복구 과정에서 기능한다. DNA리가아제 I 결핍이 초래되는 경우 면역결핍이 일어나며 DNA 손상 인자에 대한 민감도를 증가시킨니다. 10개 궤양성 대장염과 관련이 있는 유전자가 후보군으로 발견되었고, 그중 6개가 변이를 가지는 경우 질환을 더 일으키는 방향을 가지는 loci들이 발견되었다. 궤양성 대장염GWAS를 통해 잘 알려진 2개의 관련 유전자는 BTNL2, rs28362677(P = 3.17ⅹ10^(-6); OR 1.30)와 HLA-DQB1, rs1049107 및 Pⅹ1049056(P = 1.52ⅹ10^=(-6) 및 6); ^(−6) OR 1.28)가 있었다. BTNL2는 HLA 영역에 위치하며 그 전령 RNA는 장에서 가장 많이 발현된다. 결론: 대량의 환자군으로부터 얻은 염기 서열 데이터를 활용하여 염증성 장질환과 연관성 있는 단백질 기능에 영향을 미치는 변이들을 후보 변이들을 추출하고, 후보변이들이 genotype에 따라 염증성 장질환의 예후 검사로 잘 알려진 C-reactive protein, 백혈수, 혈소판 혈액 수치들에 대해도 영향을 미치는지 확인하였다. 크론병의 후보군과 궤양성 대장염의 후보군들의 생리학적 기능을 비교함으로서 염증성 장질환의 생물학적 과정을 이해하였다.

Rare-Variant Approaches to Complex Traits Across Population Biobanks

Venkataraman, Guhan Ram ProQuest Dissertations & Theses Stanford Universit 2021 해외박사(DDOD)

소속기관이 구독 중이 아닌 경우 오후 4시부터 익일 오전 9시까지 원문보기가 가능합니다.

Complex diseases are a significant global burden, accounting for 70% of deaths in the U.S. annually. For example, 70,000 new cases of inflammatory bowel disease are diagnosed every year. Many such diseases have underlying genetic etiologies responsible for their pathology. Understanding their genetic basis could lead to more timely diagnosis and improved prognosis. Furthermore, human genetics presents an opportunity to identify new therapeutic targets. However, while much of the diseasecausative common variation is well-documented, our understanding of rare, disease-contributory variation is sparse, largely due to the lack of power (limited sample size) and ascertainment to detect such variation with accuracy. While the above goal of understanding rare variation is not achievable with small cohort (n ≤ 100,000) studies, population-level biobanks (n ~ 500,000) o↵er the ability to study this type of genetic variation. These datasets present a unique opportunity for rapid discovery of robustly-validated disease-causative variants because they possess large sample sizes, better population-level annotations, and next-generation sequencing technologies like whole-exome and -genome sequencing.This dissertation contains six Chapters. In Chapter 1, I introduce key terms and methodologies and then outline my contributions to the human genetics space that are a) key and b) ancillary to the dissertation. In the subsequent Chapters (2, 3, 4, and 5), I detail the main tenets of my research, which involve the application and development of methods for the study of rare variation in the genome.The HLA region in chromosome 6 of the genome, specifically, is a hyper-polymorphic source of rare variation and of much interest because of its involvement in autoimmune disorders and cancers. Chapter 2 explores the human leukocyte antigen (HLA) region in the UK Biobank, cataloging smalland large-e↵ect rare variations that explain additional heritability of complex diseases. In addition to single-allele association testing, we also perform the Bayesian Model Averaging technique for model selection, explore non-additive associations, and investigate the e↵ect of HLA homozygosity on phenotype [184].In Chapter 3, I introduce Multiple Rare-variants and Phenotypes (MRP), a novel, flexible, Bayesian framework for rare-variant signal aggregation across variants, studies, and phenotypes. I generate gene-based results across exome data for more than 2,000 traits in the UK Biobank, v compare these findings to the existing literature, and identify novel gene-phenotype associations. In addition, I explore the use of MRP in the multi-phenotype setting by grouping related sets of biomarkers; in this joint phenotype setting, we find several genes for which power gains were substantial. This work was submitted to the American Journal of Human Genetics in 2021 [182].As a result of performing gene-based tests, the interpretability of the e↵ect profile of individual variants on the single or multivariate phenotype is not easily characterized. Chapter 4 details a corollary method to MRP, the Multiple Rare-variants and Phenotypes Mixture Model (MRPMM), which clusters rare variants into groups based on their e↵ects on a multivariate phenotype. I apply this method, as in MRP, across all single traits in the UK Biobank as well as lipid-related and renal-related multivariate phenotypes. This work was submitted to PLOS Genetics in 2021 [181].While the previous Chapters focus on applying methods across population biobanks, I also leverage targeted studies to perform detailed analysis of single phenotypes. In Chapter 5, I, along with collaborators from the International Inflammatory Bowel Disease Genetics Consortium, identify rare coding variants newly associated with Crohn’s disease using a mixed-model approach with a software called SAIGE. We submitted this work to Nature Genetics in 2021 [158].I conclude in Chapter 6 with a summary of the value in studying rare variation in the genome, the takeaways from my research, and the areas in which future research should go.

Interaction analyses of obesity traits using Korean cohorts and UK Biobank data

이원준 경희대학교 대학원 2020 국내석사

Obesity results from alteration of balance between calorie intake and expenditure and leads to lifestyle-related diseases. Although genome-wide association studies (GWASs) have revealed many obesity-related genetic factors, their interactions with calorie intake and smoking status remain unknown. This study is divided into two parts. In first part, we aimed to investigate interactions between calorie intake and the polygenic risk score (PRS) of obesity. Three Korean cohorts, Korea Association REsource (KARE; N=8,736), CArdioVAscular disease association Study (CAVAS; N=9,334), and Health EXAminee (HEXA; N=28,445) were recruited for this study. Obesity-related genetic loci were selected from previous GWASs, and two polygenic risk scores, PRS and association PRS (aPRS), were used; the former was determined from 201 SNPs from 5 GWAS data, and the latter was determined from 78 SNPs potentially associated in three Korean cohorts (meta-analysis P < 0.01). PRS and aPRS were significantly associated with BMI in all three cohorts, but PRS and aPRS did not show significant interactions with total calorie intake on BMI. Similar results were obtained for obesity. PRS and aPRS were significantly associated with obesity, but PRS and aPRS did not have significant interaction with calorie intake on obesity. We further analyzed the interaction with protein, fat, and carbohydrate intake. The results were similar to total calorie intake, and PRS and aPRS were not associated with the interaction of any of the three nutrition components. The second part, we aimed to discover novel SNPs associated with obesity traits through interaction with smoking status. UK Biobank data (N=334,808) was recruited for this study. We discovered 2 novel SNPs associated with four obesity traits (BMI, severe obesity, waist circumferences, WHR) through interaction with smoking status. We also conducted stratified analysis, we found that the effect of SNPs on obesity trats are different in smoker group and nonsmoker group. 비만은 칼로리 섭취와 소비 사이의 균형에 변화로부터 유래되며 생활과 직결된 질병으로 이어진다. GWAS는 많은 비만 관련 유전적 요인을 밝혀냈지만, 칼로리 섭취와 흡연 상태와의 상호작용은 여전히 알려지지 않고 있다. 이 연구는 두 부분으로 나뉘며 1부에서는 영양소 섭취와 비만의 polygenic risk score(PRS) 사이의 상호작용을 조사하는 것을 목표로 하였다. 이 연구에는 Korea Association REsource (KARE; N=8,736), CArdioVAscular disease association Study (CAVAS; N=9,334)와 Health EXAminee (HEXA; N=28,445) 등 3명의 한국 코호트가 사용됐다. 이전 GWAS에서 비만 관련 유전자좌가 선택되었고, PRS와 association PRS(aPRS)라는 PRS 지표가 2개가 사용되었으며, 전자는 5개의 GWAS 데이터로부터 얻은 201개의 SNP로 결정되었으며, 후자는 3개의 한국 코호트에서 연관 가능성이 있는 78개의 SNP로 결정되었다 (Meta-analysis P-value < 0.01). PRS와 aPRS는 세 가지 코호트에서 모두 BMI와 유의하게 연관되었지만 PRS와 aPRS는 BMI의 총 칼로리 섭취와 유의한 상호작용을 보이지 않았다. 또한 비만에 대해서도 유사한 결과를 얻었다. PRS와 aPRS는 비만과 상당히 관련이 있었지만, PRS와 aPRS는 비만에 대한 칼로리 섭취와 큰 상호작용을 하지 않았다. 이후 단백질, 지방, 탄수화물 섭취와의 상호작용을 추가로 분석했으며 결과는 총 칼로리 섭취와 비슷했다. PRS와 aPRS는 세 가지 영양 성분 중 어느 것 하나와의 상호작용과는 관련이 없었다. 두 번째 연구에서는 흡연 여부와의 상호작용을 통해 비만 형질과 관련된 새로운 SNP를 발견하는 것을 목표로 했다. UK Biobank data (N=334,808)가 이 연구를 위해 모집되었다. 흡연 여부와의 상호작용을 통해 4가지 비만 특성 (BMI, 중증 비만, 허리둘레, WHR)과 관련된 2개의 새로운 SNP를 발견했다. 또한 stratified analysis을 실시하여 각 SNP들이 비만 형질에 미치는 영향이 흡연자 그룹과 비흡연자 그룹에서 다르다는 것을 발견했다.

Fine-Mapping Complex Traits in Large-Scale Biobanks Across Diverse Populations

Kanai, Masahiro Harvard University ProQuest Dissertations & Theses 2022 해외박사(DDOD)

소속기관이 구독 중이 아닌 경우 오후 4시부터 익일 오전 9시까지 원문보기가 가능합니다.

Identifying causal variants for complex traits is a major goal of human genetics research. Despite the great success of genome-wide association studies (GWAS) in locus discovery, individual causal variants in associated loci remain largely unresolved, limiting the biological inference possible from follow-up experimentation. In this dissertation, I present our fine-mapping analyses of complex traits in large-scale biobanks across diverse populations to create an atlas of causal variants.We first fine-mapped complex traits using 361,194 European individuals from UK Biobank (UKBB) and gene expression using 49 tissues from GTEx (Chapter 1). We then extended our fine-mapping of complex traits to multiple populations, using 178,726 Japanese individuals from BioBank Japan and 271,341 Finnish individuals from FinnGen (Chapter 2). In total, we identified 4,518 variant-trait pairs with high posterior probability (> 0.9) of causality across the three populations. Aggregating data across populations enabled replication of 285 high-confidence variant-trait pairs as well as identification of 1,492 unique fine-mapped coding variants and 176 genes in which multiple independent coding variants influence the same trait. These results demonstrate that fine-mapping in diverse populations enables novel insights into the biology of complex traits by pinpointing high-confidence causal variants for further characterization.Next, we investigated fine-mapping accuracy in GWAS meta-analysis (Chapter 3). We demonstrated that meta-analysis fine-mapping is substantially miscalibrated in simulations and proposed a novel quality-control method, SLALOM, that identifies suspicious loci for meta-analysis fine-mapping. Having validated SLALOM performance in simulations, we found widespread suspicious patterns in existing GWAS significant loci that call into question fine-mapping accuracy. We thus urge extreme caution when interpreting fine-mapping results from meta-analysis.Finally, we introduce a new polygenic risk score (PRS) method, PolyPred, that improves cross-population polygenic prediction by combining a new fine-mapping-based predictor and a published BOLT-LMM predictor (Chapter 4). Leveraging estimated causal effects from fine-mapping enabled higher PRS transferability in non-European populations, achieving up to +32% improvement in prediction accuracy vs. BOLT-LMM using UKBB Africans.Altogether, this work demonstrates key advances in fine-mapping complex traits across diverse populations and provides insights into further variant characterization as well as improved polygenic prediction based on fine-mapping.

Dissecting Complex Disease Pleiotropy Through Multi-Trait Association Studies

Bone, William University of Pennsylvania ProQuest Dissertations 2023 해외박사(DDOD)

소속기관이 구독 중이 아닌 경우 오후 4시부터 익일 오전 9시까지 원문보기가 가능합니다.

The success of biobanks in collecting phenotype and genotype data from millions of people has dramatically changed the scale of genetic associations studies. The collection of these data has made it possible to study the genetics of many more human traits in larger cohorts, and has shown that pleiotropy, the phenomenon where a single genetic locus has an effect on multiple traits, is ubiquitous in the human genome. Pleiotropy is particularly common between cardiometabolic traits and complex diseases, such as circulating lipid levels and coronary artery disease. We can study pleiotropy to better understand the relationships between these traits and detect novel therapeutic opportunities for these diseases. Using a number of different methods, I worked to first detect pleiotropic loci that involved cardiometabolic traits and then understand the genetic mechanisms behind these pleiotropic loci. I did this by using multi-trait genetic association methods to detect loci associated with multiple traits, both in common variants via multi-trait genome-wide association studies (GWAS) and in rare variants using multi-trait gene burden analyses. A vital tool for identifying candidate causal genes at pleiotropic loci identified from multi-trait GWAS was genetic colocalization analysis between GWAS signals and expression quantitative trait loci (eQTL) and splicing quantitative trait loci (sQTL). These analyses allowed us to identify which eQTL and sQTL signals for genes have evidence of sharing the same causal variants as the GWAS signals. Part of the work presented here is the development of a framework for performing these QTL-GWAS colocalization analyses at scale. Throughout these analyses we detected several loci with evidence of pleiotropy and identify many candidate causal genes supported by statistical genetics work as well as functional work. Some of these genes, such as DOCK4 and PCSK6, may be good candidates for therapeutic targets to treat multiple diseases. These experiments show how we can use large-scale genetic and phenotypic data from biobanks to better understand the relationships between human diseases and leverage this to identify potential therapeutic targets. Supplemental files for this document include: Supplementary Methods, Supplementary Tables 1-15, and Supplementary Figures S1-S6 for Chapter 3, Supplementary Methods, Supplementary Tables 1-8, and Supplementary Figures S1-S36 for Chapter 4, and Supplementary Tables 1-3 for Chapter 5.