Genomic Reconstruction of Dietary History in Retrospective Clinical Research

Abstract

Diet is fundamental to human health, particularly in the context of disease. Adequate dietary intake is critical for managing disease progression and optimizing treatment response in hospitalized patients, but its objective assessment remains difficult. Current standard practices primarily rely on anthropometric signs such as weight loss for nutritional evaluation, yet these represent delayed physiological manifestations of dietary risks that have already occurred. Conventional methods for dietary monitoring, such as dietary recall surveys, require active patient participation and often suffer from compromised reliability. This is due to the high cognitive demand on patients’ memory and literacy at a time when they are burdened by underlying disease and hospitalization stress. To address these limitations, we applied FoodSeq, a fecal DNA sequencing method analyzing plant and animal DNA, to patients undergoing hematopoietic stem cell transplantation (HCT). In Chapter 2 of this dissertation, I present our work benchmarking FoodSeq against 150 hospital foods to confirm its ability to identify commonly consumed taxa. In Chapter 3, we demonstrate our application of FoodSeq to 878 longitudinal samples from 296 HCT patients at two tertiary centers, which provided an objective measure of oral food consumption, whose early decline within days of HCT predicted delayed platelet engraftment and reduced long-term survival in allogeneic recipients. Finally, Chapter 4 elucidates correlations between FoodSeq-measured oral food consumption and protective microbiome features, highlighting improved prognostic accuracy of mortality risk when diet and microbiome factors were modeled together. These findings underscore early nutritional recovery as key to HCT success and establish FoodSeq as a valuable tool for objective dietary monitoring in clinical settings.