Correlated factor models for exploratory analysis of complex multimodal study designs

/Abstract Most human diseases are complex, manifesting from an interplay between genes and environment over the lifespan that involve myriad biological processes. Genome-wide association studies have primarily implicated non-coding variation that is thought to lead to disease via disruption of complex, multi-level biological systems. Thus, improvements in our understanding of these fundamental processes underlying disease necessitates studying the relationship between multiple omics (multi-omic) modalities, both longitudinally and in conjunction with non-omic data. While recent years have seen an explosion of studies collecting multi-omic data in human populations, analysis of these data remains challenging both statistically and computationally. Here, I propose several new methods based on correlated latent factor models that will extend the capabilities of multi-omic inference methods to more complex study designs. I will develop model-based imputation methods that allow robust handling of missing data, enabling larger-scale studies of multi-omic biological contexts, and allowing researchers to design targeted multi-omic panels to extract the maximum amount of clinically-relevant information. I will develop multi-omic analysis methods that integrate across tissues and time points, enabling the study of dynamic molecular process and detection of systems-level impacts of intervention or disease onset. Finally, I will develop integration methods based on non-linear representation learning. This will enable detection of complex relationships between omics methods and integration with structured non-omics data such as doctor’s notes and radiographic images. To demonstrate the broad utility of the proposed methods, I will conduct collaborative analyses of varied cohorts. These include a population of individuals with subclinical atherosclerosis (MESA), a study anlyzing the relationship between microbiome features and immune health in the context of the COVID-19 pandemic (ImmunoMicrobiome), and a study of the impact of Alzheimer’s disease on neuroimaging and spinal uid biomarkers (ADNI). Completion of this research program will provide new insights into the fundamental biological processes underlying a host of common conditions, while bootstrapping the larger multi-omics research community by providing new tools that can handle complex study designs and integration tasks. Project Number: 1R35GM162151-01 | Fiscal Year: 2026 | NIH Institute/Center: National Institute of General Medical Sciences (NIGMS) | Principal Investigator: Brielin Brown | Institution: UNIVERSITY OF PENNSYLVANIA, PHILADELPHIA, PA | Award Amount: $406,250 | Activity Code: R35 | Study Section: Special Emphasis Panel[ZRG1 MCST-G (56)] View on NIH RePORTER: https://reporter.nih.gov/project-details/1R35GM16215101