Multiomic Body Mass Index signatures in blood reveal clinically relevant population heterogeneity and variable responses to a healthy lifestyle intervention

Abstract

Multiomic profiling is useful in characterizing heterogeneity of both health and disease states. Obesity exerts profound metabolic perturbation in individuals and is a risk factor for multiple chronic diseases. Here, we report a global atlas of cross-sectional and longitudinal changes associated with Body Mass Index (BMI) across 1,100+ blood analytes, as well as their correspondence to host genome and fecal microbiome composition, from a cohort of 1,277 individuals enrolled in a wellness program. Machine learning-based models predicting BMI from blood multiomics captured heterogeneous states of both metabolic and gut microbiome health better than classically measured BMI, suggesting that multiomic data can provide deeper insight into host physiology. Moreover, longitudinal analyses identified variable trajectories of BMI in response to a lifestyle intervention, depending on the analyzed omics platform; metabolomics-based BMI decreased to a greater extent than actual BMI, while proteomics-based BMI exhibited greater resistance. Our analysis further elucidated blood analyte-analyte associations which were significantly modified by obesity and partially reversed in the metabolically obese population through the program. Altogether, our findings provide an atlas of the gradual blood perturbations accompanying obesity and serve as a valuable resource for robustly characterizing metabolic health and identifying actionable targets for obesity.

Longitudinal multi-omics study reveals common etiology underlying association between plasma proteome and BMI trajectories in adolescent and young adult twins

“…Including two cohorts of different ages and populations also empowered the identification of biomolecules likely to be stable across generations and populations. This is in line with two longitudinal multi-omics studies that identified stable biomolecules in response to interventions and investigated proteome resistance to weight change [18,19]. Although BMI measurements at blood sampling were similar between the two cohorts (Table 1), both the length of follow-up and the intensity of BMI changes between the two cohorts differed.…”

“…Overall, only a few multi-omics studies quantified biomolecules (i.e., metabolites or proteins) and omics stability in relation to participants' post-intervention weight changes [18][19][20].…”

A framework towards digital twins for type 2 diabetes

“…There are hurdles associated with translating omic technologies to clinical useincluding cost, technology development, and education (38)-but our results show that their inclusion increases predictive value. Other studies have shown that omic data integrated with clinical labs can yield increased model performance, but these results are for health metrics (15) and not clinically validated tests. Second, the ability to interpret model predictions via a knowledge graph provides a robust framework for the implementation of biomedical digital twins.…”

“…In this study we used the Arivale dataset, described in detail in (14,15). Briefly, the Arivale dataset includes longitudinal data from ∼5,000 deeply phenotyped individuals undergoing a wellness program.…”

“…These data include comprehensive self-reported data coupled with multiomic data (proteomics, metabolomics, clinical labs). The process of collecting multiomic and clinical data in Arivale has been described in the existing literature (15).…”