Research

Human Microbiome

The human microbiome is an immense ecosystem of hundreds of microbial species, each with diverse and unique metabolic capabilities, living throughout our body (on our skin, in the nasal cavity, oral cavity, vagina, human milk, and of course, our intestines). The microbiome can perform tasks that the human host sometimes cannot, such as breaking down human milk oligosaccharides, which are complex sugars in human milk, that are required for healthy early-life development.

The microbiome is sensitive to both environmental exposures and host physiology. Consequently, it has been found to be a strong biomarker for many complex diseases, often predicting host phenotypes better than host genetics. We find the human microbiome particularly interesting due to two important factors: it is dynamic and modifiable, making it a prime target for biomarker discovery and therapy.

Our Approach

Unlike our DNA, the microbiome is dynamic and subject to frequent changes. Our research explores how the microbiome evolves over time and how these changes can indicate health and disease. Our toolkit includes tensor factorization, time-series analysis, and machine learning. We develop  computational methods designed to serve the entire microbiome community and apply these methods to decipher the microbiome's role in fertility, pregnancy, and lactation.

Computational Methods

Studying the human microbiome presents several challenges: (1) microbial communities are represented by noisy, high-dimensional metagenomic data, (2) they are dynamic and subject to frequent changes, and (3) they are complex, composed of diverse species with unique metabolic functions that require the integration of multiple data types. To address these challenges and identify robust, reproducible patterns that differentiate health-related phenotypes, we develop computational methods that perform dimensionality reduction and capture the ecological and eco-evolutionary dynamics of microbial communities.

Fertility

We study how the vaginal ecosystem influence reproductive health and outcomes. By profiling the microbiomes of individuals undergoing fertility treatments, we aim to identify specific microbial patterns that correlate with successful conception and pregnancy. Our approach integrates high-dimensional metagenomic data and advanced computational methods to uncover the ecological dynamics of the microbiome, allowing us to explore how these communities interact with host physiology. Ultimately, our findings seek to inform potential therapeutic strategies that could enhance fertility and improve reproductive health.

Pregnancy

We study how maternal microbial communities change during pregnancy and how these changes can be leveraged to predict pregnancy-related disorders and elucidate their underlying mechanisms.

Lactation

We study how the infant microbiome develops in early life, in both term and preterm infants, with a special focus on the interaction between human milk composition, microbial development, and the developmental origins of disease.

Human Milk

Milk is a complex, dynamic fluid essential for the growth and development of mammalian young, shaped by millions of years of evolution to support postnatal development. Human milk, often considered a "live tissue," contains thousands of signaling molecules, including nutrients, metabolites, bioactive proteins, immune factors, and microbiota. Its composition varies between mothers and changes throughout lactation.

Despite its crucial role, our understanding of human milk remains limited—remarkably, we know more about what’s in a strawberry than what’s in human milk. Traditional human milk research often focuses on single, primarily nutritive components. Although useful, this overly simplistic approach overlooks the complex systems biology and chronobiology of human milk, underestimating the importance of the interactions between its various components. Further, human milk does not function in isolation; it is part of a co-adapting system involving maternal physiology, milk composition, and infant physiology—collectively known as the “mother-milk-infant” triad—where variations in each component can influence health and developmental outcomes.

Our Approach

We employ a novel systems biology approach that situates human milk within the broader “mother-milk-infant” triad. Our approach involves in-depth molecular characterization of both nutritive and non-nutritive components of human milk, combined with the development of bespoke models that integrate community ecology theory and machine learning. This approach allows us to create holistic, low-dimensional representations of human milk as a dynamic system, which we use to predict and infer causal relationships between milk composition, its temporal dynamics, and associated health outcomes.

Visonary AI

Preeclampsia is a vascular disorder of pregnancy that can lead to multi-system organ failure. Currently, there is no reliable early diagnostic test for preeclampsia, as symptoms typically emerge only in the second half of pregnancy. Furthermore, pregnancy is considered a "window to future vascular health," with conditions like preeclampsia linked to long-term risks such as hypertension, strokes, heart disease, and vascular dementia.

Our research integrates high-resolution retinal imaging with bespoke AI model to predict adverse pregnancy outcomes, specifically focusing on the early detection of hypertensive disorders, including preeclampsia. By using the eye as a proxy for pregnancy health, this approach facilitates early risk stratification and opens new avenues for preventative care and diagnostics. 

Our Approach

We combine high-resolution, non-invasive retinal imaging with advanced AI to predict preeclampsia early in pregnancy. Our approach is based on the observation that women who will develop preeclampsia exhibit increased vascular reactivity long before symptoms appear, as placental and maternal vascular changes occur as early as the first trimester. By focusing on the retinal vasculature as a window into vascular health, we leverage these early changes for novel diagnostics. Our findings show that retinal features can accurately predict preeclampsia in the first trimester, well before clinical symptoms emerge—a capability unmatched by any existing test.

©2024 Shenhav Lab. All Rights Reserved.