Metabolomic and genomic prediction of common diseases in 477,706 participants in three national biobanks

Abstract

Identifying individuals at high risk of chronic diseases via easily measured biomarkers could improve public health efforts to prevent avoidable illness and death. Here we present nuclear magnetic resonance blood metabolomics from half a million samples from three national biobanks. We built metabolomic risk scores that identify a high-risk group for each of 12 diseases that cause the most morbidity in high-income countries and show consistent cross-biobank replication of the relative risk of disease for these groups. We show that these metabolomic risk scores are more strongly associated with future disease onset than polygenic scores for most of these diseases. In a subset of 18,000 individuals with metabolomic biomarkers measured at two time points we show that people whose scores change have dramatically different future risk of disease, suggesting that repeat measurements capture the benefits of lifestyle change. We show cross-biobank calibration of our scores. Since metabolomics can be measured from a standard blood sample, we propose such tests can be feasibly implemented today in preventative health programs.

Serum metabolomics improves risk stratification for incident heart failure

“…Screening strategies relying on such heterogeneous data sources necessarily involve employment of highly trained healthcare individuals, are difficult to standardize, expensive to collect, and as a result less effectively scalable. Third, as shown in recent approaches, [9][10][11] serum metabolomics simultaneously inform on multi-disease risk for several entities other than HF. Whilst disease-specific benchmarking against the clinical state-of-the-art is warranted, simultaneous multi-disease prediction might further expand the utility of serum metabolomics.…”

Cardiovascular risk prediction using metabolomic biomarkers and polygenic risk scores: A cohort study and modelling analyses

“…These results highlight the complementary nature of the information capture by PRSs and NMR scores. While PRSs capture the lifetime risks due to genetics [8][9][10] , NMR scores capture part of the dynamic component of risk conferred by lifestyle and environment 23 , which act on that genetic background.…”

“…Several studies have investigated the utility of biomarkers combinations from NMR platforms to improve prediction of CVD [23][24][25][26] ; however, they have focused on multi-disease prediction, used outdated clinical risk prediction scores, and have not investigated improvements relative to clinically relevant guidelinerecommended risk thresholds.…”

Integrative machine learning approaches for predicting disease risk using multi-omics data from the UK Biobank

“…Traditional models for disease prediction have primarily focused on singular data types, often overlooking the complex interplay between various biological and demographic factors. Recent studies have shown the potential of individually using metabolomics, demographics, and genomics to predict age-related diseases and mortality [4]. These findings underscore the potential of using machine learning models to leverage high-throughput data and better understand the complex non-linear relationships between multi-omics predictors.…”