Molecular and Automated Biomarkers of Human Diet and Physiology

Abstract

Accurate measurement of human diet and physiology is essential to biomedical researchand clinical care, yet many of the most widely used monitoring methods rely on subjective assessment, introducing bias and limiting comparability across populations and settings. This dissertation develops and applies molecular and automated tools for objective measurement of human diet and physiology across contemporary, archaeological, and clinical settings. In Chapter 2, I applied FoodSeq, a DNA sequencing method for identifying dietary plants from stool, to fecal samples from 1,078 individuals across seven countries, finding that plant dietary diversity increases along the epidemiological transition but does not associate with improved cardiometabolic health, largely because ultra-processed foods contain significantly more plant species than minimally processed foods. The co-occurrence of wheat and soy emerged as an objective marker of ultra-processed food status, and a geographically restricted "Salad Bowl" pattern enriched for leafy greens was found almost exclusively among U.S. participants, with economic cost and foodborne illness risk identified as barriers to its global adoption. In Chapter 3, I applied vertebrate DNA metabarcoding to coprolites from the Cave at Rio Zape in Durango, Mexico (725–950 CE), identifying 17 vertebrate dietary taxa including five freshwater fish detected in half of all samples, establishing aquatic resources as a component of Loma San Gabriel subsistence in a riverine landscape since lost to drought. In Chapter 4, I contributed to the design and validation of an automated urine output tracking device that uses load cell sensors integrated into existing toilet infrastructure to quantify urine volume, achieving a mean absolute error of approximately 12 grams in healthy volunteer testing, and described a randomized observational study to validate the device in hospitalized hematopoietic cell transplant and heart failure patients. Together, Chapters 2 and 3 show how DNA-based methods can extend dietary inference across modern and archaeological contexts, while Chapter 4 broadens the dissertation’s scope to automated physiological monitoring in clinical care.