서브메뉴
검색
Random Covariance Models, Functional Brain Networks, and Interpretable Deep Learning for Multiomic Data Integration
Random Covariance Models, Functional Brain Networks, and Interpretable Deep Learning for Multiomic Data Integration
상세정보
- 자료유형
- 학위논문 서양
- 최종처리일시
- 20260202102946
- ISBN
- 9798293861507
- DDC
- 574
- 서명/저자
- Random Covariance Models, Functional Brain Networks, and Interpretable Deep Learning for Multiomic Data Integration
- 발행사항
- [Sl] : University of Minnesota, 2024
- 발행사항
- Ann Arbor : ProQuest Dissertations & Theses, 2024
- 형태사항
- 176 p
- 주기사항
- Source: Dissertations Abstracts International, Volume: 87-03, Section: B.
- 주기사항
- Advisor: Zhang, Lin;Safo, Sandra.
- 학위논문주기
- Thesis (Ph.D.)--University of Minnesota, 2024.
- 초록/해제
- 요약Modern advances in neuroimaging technology and high-throughput next generation molecular sequencing exemplify the overall increase in volume, variety, and complexity of high-dimensional data in contemporary neurological research. In tandem with such rapid innovation, equally innovative statistical and machine learning approaches are necessary to model associations between such complex data sources and neurological outcomes or phenotypes; and more importantly, to draw meaningful inference regarding which brain regions and molecular biomarkers drive neurological disorder and disease. Focusing specifically on functional connectivity (FC) between brain regions and multiomic molecular sequencing (e.g. genomics, proteomics, transcriptomics, etc.), three distinct but interrelated studies were conducted to advance understanding of and statistical methodology for (1) adolescent depression, (2) schizophrenia, and (3) Alzheimer's disease (AD), respectively. In (1), a frequentist bi-level (group- and participant-specific) Gaussian graphical model, referred to as a Random Covariance Model (RCM), is applied for new insight into functional characteristics of adolescent depression and suicide attempt; while in (2), a Bayesian RCM with more robust participant-specific regularization is developed, tested, and applied to uncover resting-state default mode network (DMN) dynamics in schizophrenia patients. Lastly, etiological and clinical heterogeneity both complicate diagnosis and treatment of AD, and new methods to integrate heterogeneous multiomics data for interpretable molecular biomarker analyses are needed. This motivates the development and study of a new one-step interpretable deep learning framework, jointly leveraging genomics, lipidomics, and metabolomics data, for biomarker selection and multiclass prediction in (3).
- 일반주제명
- Biostatistics
- 일반주제명
- Neurosciences
- 일반주제명
- Genetics
- 일반주제명
- Biomedical engineering
- 키워드
- Schizophrenia
- 기타저자
- University of Minnesota Biostatistics
- 기본자료저록
- Dissertations Abstracts International. 87-03B.
- 전자적 위치 및 접속
- 로그인 후 원문을 볼 수 있습니다.
MARC
008260126s2024 us c eng d■001000017356537
■00520260202102946
■006m o d
■007cr#unu||||||||
■020 ▼a9798293861507
■035 ▼a(MiAaPQ)AAI31559244
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a574
■1001 ▼aNeville, Quinton Daniel.
■24510▼aRandom Covariance Models, Functional Brain Networks, and Interpretable Deep Learning for Multiomic Data Integration
■260 ▼a[Sl]▼bUniversity of Minnesota▼c2024
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2024
■300 ▼a176 p
■500 ▼aSource: Dissertations Abstracts International, Volume: 87-03, Section: B.
■500 ▼aAdvisor: Zhang, Lin;Safo, Sandra.
■5021 ▼aThesis (Ph.D.)--University of Minnesota, 2024.
■520 ▼aModern advances in neuroimaging technology and high-throughput next generation molecular sequencing exemplify the overall increase in volume, variety, and complexity of high-dimensional data in contemporary neurological research. In tandem with such rapid innovation, equally innovative statistical and machine learning approaches are necessary to model associations between such complex data sources and neurological outcomes or phenotypes; and more importantly, to draw meaningful inference regarding which brain regions and molecular biomarkers drive neurological disorder and disease. Focusing specifically on functional connectivity (FC) between brain regions and multiomic molecular sequencing (e.g. genomics, proteomics, transcriptomics, etc.), three distinct but interrelated studies were conducted to advance understanding of and statistical methodology for (1) adolescent depression, (2) schizophrenia, and (3) Alzheimer's disease (AD), respectively. In (1), a frequentist bi-level (group- and participant-specific) Gaussian graphical model, referred to as a Random Covariance Model (RCM), is applied for new insight into functional characteristics of adolescent depression and suicide attempt; while in (2), a Bayesian RCM with more robust participant-specific regularization is developed, tested, and applied to uncover resting-state default mode network (DMN) dynamics in schizophrenia patients. Lastly, etiological and clinical heterogeneity both complicate diagnosis and treatment of AD, and new methods to integrate heterogeneous multiomics data for interpretable molecular biomarker analyses are needed. This motivates the development and study of a new one-step interpretable deep learning framework, jointly leveraging genomics, lipidomics, and metabolomics data, for biomarker selection and multiclass prediction in (3).
■590 ▼aSchool code: 0130.
■650 4▼aBiostatistics
■650 4▼aNeurosciences
■650 4▼aGenetics
■650 4▼aBiomedical engineering
■653 ▼aAdolescent depression
■653 ▼aAlzheimer's disease
■653 ▼aFmri functional connectivity
■653 ▼aMultiomic data integration
■653 ▼aRandom Covariance Model
■653 ▼aSchizophrenia
■690 ▼a0308
■690 ▼a0317
■690 ▼a0369
■690 ▼a0541
■71020▼aUniversity of Minnesota▼bBiostatistics.
■7730 ▼tDissertations Abstracts International▼g87-03B.
■790 ▼a0130
■791 ▼aPh.D.
■792 ▼a2024
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17356537▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.
