Elhanan Borenstein

Companion dogs are a valuable geroscience model because they share human environments, develop similar age-related diseases, and have compressed lifespans. The Dog Aging Project (DAP) is a large, prospective cohort designed to investigate intrinsic and extrinsic factors associated with canine aging and longevity. We analyzed 41,047 dogs from the 2024 DAP Curated Data Release with baseline Health and Life Experience Survey (HLES) data and mortality follow-up through 2024. All-cause mortality was assessed using owner-reported death dates. We estimated survival by size class, breed background (American Kennel Club-recognized vs. non-AKC/mixed-breed), and sex using Kaplan–Meier methods and Cox models. We further evaluated geographic variation and conducted a survey-wide association study (SWAS) of 804 HLES variables using Cox models stratified by demographics with false discovery rate control. Median lifespan ranged from 14 years in small dogs to less than 10 years in giant AKC-recognized breed dogs. Within size-sex strata, AKC-recognized breed dogs lived modestly shorter lives than non-AKC/mixed-breed dogs, and females had slightly longer lifespans than males. After demographic adjustment, rural-suburban-urban differences were no longer evident, but there was significant regional and state-level variation, which correlated positively with human mortality patterns ( r = 0.43, ρ = 0.63, p = 0.0017). The SWAS identified 72 baseline variables associated with mortality, primarily from Health Status, Physical Activity, Behavior, Environment, and Diet domains, with many showing clear exposure–response relationships. This initial DAP survival analysis confirms strong demographic effects on canine longevity, reveals shared geographic mortality patterns between dogs and humans, and highlights survey-based health, behavioral, and environmental factors warranting targeted causal investigation.

Background & Aims: Delayed diagnosis of inflammatory bowel disease (IBD) is associated with worse outcomes. Existing prediction studies often rely on small cohorts or specialized biomarkers that limit scalability. We evaluated whether machine learning models based on routinely collected clinical data can predict future Crohn’s disease (CD) and ulcerative colitis (UC) years before diagnosis in a large population-based cohort. Methods: We conducted a retrospective study using electronic medical records from Maccabi Health Services (Israel), including 81 960 individuals (21 695 with IBD). Longitudinal laboratory tests, body mass index, demographics, and comorbidities were used to train Explainable Boosting Machine (EBM) models. Two frameworks were examined: (1) diagnosis-anchored prediction 0.5-2 years before diagnosis, and (2) age-based screening, estimating risk at predefined ages with future prediction windows. Performance was assessed using area under the receiver operating characteristic curve (AUROC) and relative risk (RR) in high-risk subgroups. Results: In diagnosis-anchored models, AUROC for CD ranged from 0.71 to 0.75 up to 2 years before diagnosis and from 0.62 to 0.64 for UC. In age-based screening, the best performance was observed at age 25 with a 1-11-year prediction window (AUROC 0.72 for CD; 0.70 for UC). Individuals in the top decile of predicted risk had an approximately 3-fold higher risk of future IBD, with RR remaining >2.5 for CD and >2.0 for UC for shorter windows. Kaplan–Meier analyses showed clear separation between high- and low-risk groups (P < .001). Conclusions: EBM models using routine clinical data can identify individuals at increased future IBD risk, supporting scalable, interpretable population-level risk stratification and earlier evaluation.

The extraordinary lifespan variation in domestic dogs provides a natural experiment for testing how intrinsic rates of biological aging shape lifespan. Using 1640 methylomes from 894 dogs, we developed an epigenetic clock that predicted mortality and demonstrated that epigenetic aging is fastest early in life. At orthologs of human age-associated genes, dogs exhibited concordant age effects on promoter methylation, highlighting conserved remodeling of immune pathways. We found that larger and male dogs, which are shorter lived, exhibit accelerated molecular aging. Distinct epigenetic architectures mediated these effects: Sex-dependent methylation changes were concentrated on the X chromosome, whereas size-associated methylation was especially pronounced at transposable elements (TEs). These findings show that epigenetics reflects lifespan differences in dogs and identifies TEs as potential mediators of size-associated lifespan.

Motivation: A central goal in human gut microbiome research is to identify disease-associated functional shifts, an objective increasingly pursued through metagenomic and metabolomic assays. However, common differential abundance analyses of genes or metabolites often yield long and difficult-to-interpret feature lists. Aggregating features into predefined pathways can improve interpretability but relies on fixed pathway boundaries that may not reflect context-specific functional changes. Moreover, even when paired metagenomic-metabolomic data are available, they are often analyzed separately or linked only through simple statistical associations. Results: We introduce MAAMOUL, a knowledge-based computational framework that integrates metagenomic and metabolomic data to identify disease-associated, data-driven microbial metabolic modules. Leveraging prior knowledge of bacterial metabolism, MAAMOUL maps disease-association scores onto a global microbiome-wide metabolic network and identifies custom modules enriched for altered genes and metabolites. Applying MAAMOUL to inflammatory bowel disease (IBD) and irritable bowel syndrome (IBS) datasets revealed significant disease-associated modules not detected by conventional pathway-level analysis. In IBD, modules reflected disrupted sulfur and aromatic amino acid metabolism and enhanced microbial nucleotide salvage, whereas in IBS they linked purine and nicotinate/nicotinamide metabolism. These results demonstrate that network-guided multi-omic integration can uncover coherent functional shifts in the gut microbiome overlooked by single-omic or purely statistical approaches. Availability and implementation: MAAMOUL is available as an R package at https://github.com/borenstein-lab/MAAMOUL.

Companion dogs (Canis lupus familiaris) offer a unique model for studying the gut microbiome and its relation to aging due to their cohabitation with humans, sharing similar environments, diets, and healthcare practices. Here, we present the Dog Aging Project (DAP) Precision cohort, a large population-wide study of the canine gut microbiome. This cohort encompasses over 900 dogs of diverse breeds, environments, and demographics living across the United States. Coupling fecal shotgun metagenomic sequencing with phenotypic and environmental surveys and clinical lab tests, we explore the intricate relationships between microbiome composition, aging, and key factors such as health and living conditions. Our analyses identify multiple factors associated with microbiome composition, including dietary preferences such as commercial versus home cooked nutrition, and behaviors such as coprophagy (feces eating). In addition, we find age-associated gradual shifts in microbiome composition, supporting the development of a metagenomics-based population-level model for canine age prediction based on microbial signatures. We further examined which age-associated microbial patterns observed in humans are recapitulated in dogs by comparing our cohort with the Lifelines-DEEP cohort. Overall, these findings offer insights into the role the gut microbiome plays in our four-legged companions, with potential implications for veterinary medicine and translational aging research.

Background: Biomarkers are needed to predict treatment response and guide therapeutic decisions in Crohn disease (CD). We aimed to develop and validate a multi-omics machine learning (ML) model to predict response to nutritional therapy in pediatric CD. Methods: Treatment-naive children with newly diagnosed CD who were initiating exclusive enteral nutrition (EEN) were prospectively enrolled in this study. Metabolomics and lipidomics were measured in the serum and stool, as well as the fecal microbiome. Following feature selection via minimum redundancy maximum relevance, random-forest models were constructed for single- and multi-omics and performances were evaluated. The models were externally validated in an independent prospective cohort of treatment-naive children and young adults with CD treated with EEN. Results: The discovery cohort consisted of 50 children (mean ± SD age 14.3 ± 2.7 years), of whom 34 (68%) responded to EEN. Combining complementary signals from host metabolism, gut microbiota, and lipid profiles from serum and stool in a multi-omics ML model yielded a model for predicting treatment response (training accuracy 94%; 95% CI, 82%-100%). Key predictive features included serum metabolites (2-hydroxyglutaric acid, Cer[d18:0/22:0], and HexCer[d18:1/d26:1]), fecal metabolites (3-methyladipic acid, DG[16:0 20:0], PC aa C42:2), and microbial taxa (family Bifidobacteriaceae and genus CAG-56). The validation cohort consisted of 21 patients of whom 12 (57%) responded to EEN. The multi-omics model performance achieved an area under the receiver operating characteristic curve (AUROC) of 0.81 (95% CI, 0.6-1.0). Clinical and endoscopic features did not improve the predictive ability of the model. Conclusion: As a proof-of-concept, we showed that integrated multi-omics ML models can predict EEN response in pediatric CD patients, supporting their potential use in precision nutrition and personalized care strategies.

Multi-omic studies in human microbiome research hold great potential for advancing our understanding of host-microbiome interactions. However, despite the growing availability of multi-omic datasets, analysing such data remains a major conceptual, analytical and computational challenge. Introduction of new multi-omic integration methods to address these challenges further complicates researchers' efforts to navigate this expanding field. In this Review, we outline the landscape of multi-omic integration methods in the context of human microbiome research. In contrast to previous reviews, we specifically emphasize the different biological questions addressed by various integration approaches, including questions related to interactions between different molecular layers, molecular shifts that occur in disease, subgrouping of patients based on molecular profiles, and identification of biological mechanisms that underlie such associations. Our aim is to provide a timely, convenient and comprehensive resource for the microbiome research community, allowing researchers to identify the multi-omic integration approach that is best suited to their data and objectives.

Motivation: A central goal in human gut microbiome research is to identify disease-associated functional shifts, an objective increasingly pursued through metagenomic and metabolomic assays. However, common differential abundance analyses of genes or metabolites often yield long and difficult-to-interpret feature lists. Aggregating features into predefined pathways can improve interpretability but relies on fixed pathway boundaries that may not reflect context-specific functional changes. Moreover, even when paired metagenomic-metabolomic data are available, they are often analyzed separately or linked only through simple statistical associations. Results: We introduce MAAMOUL, a knowledge-based computational framework that integrates metagenomic and metabolomic data to identify disease-associated, data-driven microbial metabolic modules. Leveraging prior knowledge of bacterial metabolism, MAAMOUL maps disease-association scores onto a global microbiome-wide metabolic network and identifies custom modules enriched for altered genes and metabolites. Applying MAAMOUL to inflammatory bowel disease (IBD) and irritable bowel syndrome (IBS) datasets revealed significant disease-associated modules not detected by conventional pathway-level analysis. In IBD, modules reflected disrupted sulfur and aromatic amino acid metabolism and enhanced microbial nucleotide salvage, whereas in IBS they linked purine and nicotinate/nicotinamide metabolism. These results demonstrate that network-guided multi-omic integration can uncover coherent functional shifts in the gut microbiome overlooked by single-omic or purely statistical approaches. Availability: MAAMOUL is available as an R package at https://github.com/borenstein-lab/MAAMOUL .

Despite well-established benefits of exercise on metabolic regulation and the gut microbiome (GM), its impact on body composition is inconsistent and often attenuated by metabolic adaptation. This compensation mechanism adjusts energy expenditure including total daily energy expenditure (TDEE) and resting metabolic rate (RMR). Intra-individual variation in exercise response remains unclear, but might be explained by the GM. In this well-controlled study, we investigated the relationship between aerobic exercise, GM composition, and metabolic adaptation in a cohort of 16 sedentary overweight adults (ages 21–45, 50% female) over a 12-week moderate-intensity intervention (65–75% HRmax; 20 kcal/kg/week). Pre- and post-intervention RMR was measured via whole-room calorimetry, TDEE by doubly labeled water, and GM composition via shotgun metagenomics. While body composition did not change at the group-level, a subset of participants (“responders”) showed improved body composition and aerobic capacity. Using machine learning, we identified bacterial species, including Faecalibacterium prausnitzii species, whose abundance pre-training is predictive of response. Additionally, we found that responder GM communities are more compositionally cohesive and post-training increases in GM diversity are associated with higher TDEE and RMR. These findings highlight the complex interaction between exercise, metabolism and the GM, and suggest that baseline GM characteristics may contribute to individual variability in metabolic adaptation. This insight may help guide microbiome-informed strategies to enhance exercise efficacy. Clinical trial registration: ClinicalTrials.gov , identifier NCT04460040.

Summary: TAGINE is a feature engineering algorithm that leverages the microbial taxonomic tree to optimize feature sets in microbiome data for predictive modeling. The algorithm starts with features at high taxonomic levels and iteratively splits them into lower-level clades in cases where it improves predictive accuracy, ultimately producing a feature set spanning multiple taxonomic levels. This approach aims to markedly reduce the number of features while preserving biological relevance and interpretability. We compare TAGINE’s performance to other standard and taxonomy-based feature engineering methods on several different datasets, and show that TAGINE yields more compact feature sets and is orders of magnitude faster than other methods, while maintaining predictive accuracy. Availability and implementation: TAGINE is freely available under the MIT license with source code available at https://github.com/borenstein-lab/tagine_fe.