Back to search

Boston University

Advanced statistical methods for omics data integration

Abstract

dc:description.abstract

Biomedical studies increasingly collect multi-omics data across molecular layers, over time, and from heterogeneous cohorts. These data provide important opportunities to understand complex biological processes more comprehensively, but also introduce substantial challenges for statistical, artificial intelligence, and machine learning (AI/ML) methodologies, including population heterogeneity, multimodal data integration, and the development of generalizable and interpretable models. Motivated by applications in aging research, this dissertation develops advanced statistical and machine learning methods for omics data integration, with an emphasis on cross-modal and cross-cohort learning. In Project 1, we introduce ONDSA, a false discovery rate-controlled network testing framework for comparing multiple Gaussian graphical models. ONDSA identifies shared and differential omics network structures across biological groups and addresses a key methodological gap in error-controlled comparison of precision matrices across multiple groups. Project 2 develops MGRFusionNet, a multimodal graph-recurrent fusion architecture designed to investigate how molecular omics and longitudinal phenotypic trajectories jointly relate to future health outcomes. The proposed method captures nonlinear and time-dependent risk dynamics, as well as cross-modal interactions, and enables interpretable attribution of feature importance and interaction effects, thereby improving both prediction and interpretation. In Project 3, we propose Trans-GCR, a transfer learning method for network-structured high-dimensional data that leverages large external cohorts to improve learning in smaller target cohorts. Trans-GCR explicitly accounts for domain shift between source and target cohorts, achieves computational efficiency, and is supported by theoretical guarantees. Overall, this dissertation presents a unified framework for cross-modal and cross-cohort learning in multi-omics data integration. The proposed methods are motivated by aging-related applications but are broadly applicable to complex biomedical studies involving high-dimensional, heterogeneous, and multimodal data.

Degree

thesis:*
Name thesis:degree_name
Doctor of Philosophy
Level thesis:degree_level
doctoral
Discipline thesis:degree_discipline
Biostatistics
Grantor
Boston University
Year dc:date.issued
2026

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Chen, Jiachen
Advisor dc:contributor.advisor
  • Lunetta, Kathryn L
ORCID dc:identifier.orcid

Subjects

dc:subject × 3

Rights

Language dc:language.iso
en_US

Identifiers

dc:identifier.*
Handle dc:identifier.uri
https://hdl.handle.net/2144/53622

Chain of custody

source
Harvested from
Boston University
Base URL
open.bu.edu/oai/request
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
citation

Chen, Jiachen. Advanced statistical methods for omics data integration. doctoral thesis, Boston University, 2026. https://hdl.handle.net/2144/53622