Nonlinear and Multivariate Causal Analysis of Diverse Aging Populations

Metabolic diseases, including obesity and type 2 diabetes, are complex, multifactorial conditions caused by the interplay between genetic background, environmental exposures, and behavioral patterns. Understanding how these factors interact to influence metabolic health throughout the lifespan requires an integrated approach that combines molecular, physiological, and behavioral data. During my PhD research project, I applied a systems biology framework to study gene-environment (GxE) interactions underlying metabolic phenotypes in mice. My research encompasses three complementary projects: (1) analysis of proteomic data from targeted tissues associated with metabolic diseases, (2) intervention studies combining dietary restriction with targeted treatments (nicotinamide riboside (NR)) to mitigate aging, and (3) longitudinal behavioral monitoring using automated home cage systems.
Using a large and genetically diverse mouse population (n = 2,157), stratified by age, diet (Chow (CD) vs. High-Fat (HF)), and strain, I conducted proteomic analysis on white adipose tissue (WAT) samples from 191 individuals. WAT is a metabolically active tissue that plays a central role in energy balance and inflammation. Proteomic profiling has revealed distinct molecular signatures associated with dietary interventions and age. These changes included modifications in lipid metabolism, mitochondrial function, and inflammatory signaling. The findings provided a direct readout of the biological consequences of chronic HF diet exposure and aging, reinforcing the value of proteomics in capturing functional molecular changes. To enhance the interpretation and prediction of metabolic outcomes, I applied advanced machine learning (ML) models to analyze and predict various phenotypes, such as Body Weight, based on WAT proteomic data. This approach allowed me to systematically integrate and synthesize information derived from the intricate tissue proteome alongside corresponding phenotypic data, enabling a more comprehensive understanding of the underlying biological functions and their implications for metabolism. ML-based feature selection (FS) methods revealed key proteins and pathways that were consistently associated with Body Weight and Diet. These models enabled robust prediction of metabolic phenotypes and highlighted the added value of using ML to uncover patterns not apparent through traditional statistical analyses. Importantly, the selected features offered insight into potential biomarkers and therapeutic targets relevant to metabolic health. In a related aspect of this research, the collaborative project with Leibniz Institute on Aging, I studied the combined effects of NR supplementation and late-life dietary restriction (DR) in aged mice by analyzing transcriptomic and proteomic data from the liver. Overall, the results indicate that NR supplementation enhances the benefits of late-life dietary restriction by improving metabolic resilience, hematopoietic stem cell (HSC) function, and proteostasis, ultimately contributing to an extended lifespan in aged mice.
Finally, I validated the use of the Digital Ventilated Cage (DVC) system for high-throughput, non-invasive home cage monitoring (HCM) of spontaneous activity in three BXD female mouse strains: BXD40, BXD43, and BXD100. These strains have previously been shown to exhibit a tendency toward voluntary exercise behavior. The DVC system effectively captured daily activity patterns and behavioral responses to environmental stimuli. The analysis of the DVC system data revealed a decline in activity as age increased, highlighting the biological effects of aging and environmental adaptation. This study serves as a pilot project to validate the DVC system and to compare its outcomes with those of another study conducted on running mice.
These projects operate together to enhance our understanding of how metabolic traits are influenced by GxE interactions. They provide comprehensive datasets for exploring behavior, physiology, and molecular biology. The findings have significant implications for biomarker discovery, precision nutrition, and the creation of personalized interventions aimed at promoting metabolic health and healthy aging.