Background
Obesity drives a plethora of metabolic alterations and is a risk factor across a wide range of chronic diseases. Body Mass Index (BMI) can determine obesity at the population level, but cannot alone stratify heterogeneous metabolic health states. Multiomic profiling in blood, however, has the potential to reveal population heterogeneity for both health and disease states. A study from the Institute for Systems Biology in Seattle looked at cross-sectional and longitudinal changes in >1,100 blood analytes related to variation in BMI in ~1,300 individuals enrolled into a wellness program (Arivale). The data collected included genomics, metabolomics, gut microbiome composition, anthropometric measurements and plasma proteomics using the Olink® Target 96 CVD II, CVD III and Inflammation panels.
Outcome
Overall, plasma multiomics captured 48–78% of the total variance in BMI. The study design allowed facilitated the investigation of the similarities and differences between omics platforms according to the physiological health state of each individual across the BMI spectrum. Using data from all analytes, machine learning (ML) models were trained to predict baseline BMI for each of the omics platforms alone and in combinations. In contrast to a model including obesity-related standard clinical measures, all of the omics-based models demonstrated significantly higher performance in BMI prediction (confirmed in an independent validation cohort). Among individual datasets, proteomics provided the best prediction (marginally better than the metabolomics data), but the combined omics dataset was superior to any individual measurements, suggesting that the different omics technologies provide complementary information.
The ML-derived predictive modeling in the longitudinal analyses identified variable BMI trajectories for different omics measures in response to a healthy lifestyle intervention; metabolomics-inferred BMI decreased to a greater extent than actual BMI, whereas proteomics-inferred BMI exhibited greater resistance to change. As well as reflecting mechanistic information related to obesity, they also suggested that the omics-based models might indicate the early transition toward clinical manifestations of obesity-related chronic diseases. The authors concluded that this data could provide a valuable resource for characterizing metabolic health and identifying actionable targets for health management.