Reconstruction of Individualized, Time-Evolving Immunometabolic Systems in Critical Illness with In Silico Causal Simulation Restricted; Files & ToC
Kobara, Seibi (Spring 2026)
Abstract
Critical illness represents a complex immunometabolic dyshomeostasis, contributing to substantial global mortality and long-term sequelae. Beyond survival, survivors of critical illness frequently experience persistent impairments, including reduced physical function.
The intermediate immunometabolic states that govern these outcomes evolve rapidly. Furthermore, because identical molecular expression levels can arise from distinct biological contexts, such measurements are most informative when interpreted in relation to the processes that produced them.
However, the temporal evolution of individual immunometabolic states and their contextual determinants remains underexplored. As a result, optimal intervention targets, dosing strategies, and critical windows for intervention remain poorly defined.
The overarching objective of this dissertation is to develop patient-specific, time-resolved systems models capable of identifying molecular targets, intervention timing, and dosing strategies that can shift individuals away from adverse clinical trajectories.
Pre-existing comorbidities are well-established risk factors and represent a static baseline context associated with increased mortality in critical illness. However, associations between comorbidity and post–critical illness physical functioning are incompletely characterized due to non-survivor censoring. Bias-adjusted analyses revealed no independent association between comorbidity and physical impairment, indicating that static baseline comorbidity alone does not fully account for post–critical illness functional outcomes and motivating consideration of intermediate, time-evolving biological processes.
To address dynamic biological states and their interactions over time, this dissertation introduces HARBOR, a Bayesian causal inference framework that estimates temporal molecular interaction networks while conditioning on time-varying covariates. HARBOR enables robust inference of dynamic molecular relationships in the presence of substantial missingness and supports counterfactual evaluation of candidate molecular interventions and dosing strategies. However, temporal molecular models typically assume homogeneous intracellular regulatory machinery across individuals. To address this limitation, we develop BIOCURRENT, a donor-specific intracellular pseudotime geometry inference framework that captures heterogeneity in intracellular regulatory states.
In summary, this dissertation presents a set of computational frameworks for modeling temporal-context–aware immunometabolic interactions in longitudinal multi-omics data and intracellular geometry in single-cell transcriptomic data. Together, these approaches enable patient-specific evaluation of molecular regulatory systems, providing a principled foundation for precision intervention modeling in critical illness.
Table of Contents
This table of contents is under embargo until 27 May 2027
About this Dissertation
| School | |
|---|---|
| Department | |
| Degree | |
| Submission | |
| Language |
|
| Research Field | |
| Keyword | |
| Committee Chair / Thesis Advisor | |
| Committee Members |
Primary PDF
| Thumbnail | Title | Date Uploaded | Actions |
|---|---|---|---|
|
|
File download under embargo until 27 May 2027 | 2026-03-31 16:39:37 -0400 | File download under embargo until 27 May 2027 |
Supplemental Files
| Thumbnail | Title | Date Uploaded | Actions |
|---|