Dr. Arnold and Dr. Young are part of a multi-institutional team receiving support through the National Institutes of Health’s (NIH) Computational Modeling of Hormone Homeostasis Initiative.
U-M BME Associate Professor Kelly Arnold, along with William Henry Fitzbutler Collegiate Professor of Internal Medicine/Infectious Disease Vincent Young, and collaborators will create mechanistic microbiome “digital twins” to help researchers predict how hormonal differences influence therapeutic treatment response.
Dr. Arnold and Dr. Young are part of a multi-institutional team receiving support through the National Institutes of Health’s (NIH) Computational Modeling of Hormone Homeostasis Initiative, with a goal of helping to develop a new generation of computational tools to model how hormones interact with microbial communities and influence therapeutic responses across individuals. The three-year project, NIH Award 1OT2OD042817, is led by principal investigator Jason Papin of the University of Virginia and includes additional collaborators from U-M, UVA, Johns Hopkins University, University of Connecticut, and University of Illinois-Chicago.
The initiative is a joint effort of the NIH Office of Research on Women’s Health and the Division of Program Coordination, Planning, and Strategic Initiatives, with funding for this award provided by NIH’s newly established Office of Research Innovation, Validation, and Application. Its first awards total $21 million, distributed across 6 awards, and are intended to expand national capacity for developing human-based, data-driven models of hormone-related health.
Hormone activity can vary with sex, age, reproductive stage, medication use and other individual factors. These variations may affect how people respond to drugs, including the dose needed for a treatment to be effective and the likelihood of side effects. Existing mathematical models, however, often do not adequately capture these differences.
The new initiative seeks to address that gap by supporting in silico—or computer-based—models that use human biological data to simulate hormone activity across the life course, from adolescence through older age.
“The idea is that this modeling framework will help us understand how hormones contribute to therapeutic responses,” Dr. Arnold said. “That could include differences between women and men, pre- and postmenopausal women, women who use hormonal contraceptives and those who do not, or younger and older men.”
Dr. Arnold’s laboratory studies the microbiome: the communities of microorganisms that live throughout the human body, particularly on mucosal surfaces. Her team has focused extensively on the vaginal microbiome and how changes within its microbial community may contribute to disease or susceptibility to disease.
The laboratory uses mechanistic computational models based on ordinary differential equations, a mathematical approach long used by engineers to describe dynamic systems. In this case, the models represent interactions among bacterial species, including how they compete, cooperate and shape the behavior of the microbial community as a whole.
“Once the community is represented computationally, it becomes a valuable framework for testing hypotheses and understanding the role of individual species,” Dr. Arnold said. “It also allows us to examine why changes in microbial species may—or may not—contribute to differences in a person’s response to a therapy or drug.”
That variability is particularly evident with probiotics, which introduce bacteria believed to be beneficial into an existing microbial community. While probiotics can be helpful for some people, they may have little effect for others.
“We often select a species that we think is beneficial and introduce it into the community, but we do not always know why it helps some people and not others,” Arnold said. “These models could give us a computational framework for testing those interventions on a personalized basis.” Likewise the same framework could be valuable for assessing differences in antibiotic responses.
As part of the NIH-supported project, the U-M team will work to create personalized microbiome models using measurements of an individual’s microbial composition. Artificial intelligence tools will help generate mechanistic equations that describe the person’s microbial community and its interactions.
Dr. Arnold describes the concept as a microbiome “digital twin”—a virtual representation that researchers could use to simulate biological changes and potential interventions.
“This is similar to a microbiome digital twin, but it is mechanistic,” Dr. Arnold said. “Rather than asking AI to create a black box of relationships, we are incorporating the underlying biological mechanisms. AI helps us write the equations, but we can still understand what is happening behind the model.”
The team plans to test the framework in two health contexts: bacterial vaginosis and metabolic dysfunction-associated steatotic liver disease (MASLD), both of which involve complex relationships among hormones, metabolism, microbial activity and individual biological differences. Although those conditions will serve as initial test cases, the researchers intend to develop a broadly applicable tool.
The work brings together expertise across multiple biological scales. Dr. Arnold’s group will contribute computational models of microbial communities, while U-M microbiome researcher Dr. Young and his laboratory will provide experimental microbiology expertise. Dr. Papin’s group at UVA will contribute metabolic modeling, and other researchers at UVA, Johns Hopkins, and University of Connecticut will help integrate models that operate at different scales.
A Johns Hopkins collaborator, Associate Professor of BME Felim Mac Gabhann, will develop pharmacokinetic models describing how hormones are distributed and metabolized throughout the body. These models could help researchers examine why hormone concentrations and effects differ among the blood, liver, gut, vaginal mucosa and other tissues.
“If we connect the community models with metabolic models, we can begin to understand the metabolic events that are driving changes in the microbiome,” Dr. Arnold said. “By combining that information with models of hormone distribution throughout the body, we hope to build a platform that is useful across many different scenarios.”
The researchers ultimately plan to make the modeling platform freely available so other scientists can adapt it to new research questions, populations and therapeutic applications.
By integrating hormonal biology, microbial ecology, metabolism and pharmacokinetics, the project could strengthen researchers’ ability to predict individualized treatment responses before moving into clinical testing. It also may help scientists design studies that better reflect biological variation in real-world populations.
The broader NIH initiative aims to advance precision medicine while ensuring that biological differences are incorporated into the study of hormone regulation. The resulting tools could support more informative treatment evaluation, stronger research studies and better clinical decisions for people and communities.