Unsupervised Learning-Derived Complex Metabolic Signatures Refine Cardiometabolic Risk

Area of Science:

• Cardiovascular Science
• Metabolomics
• Genetics

Background:

• Cardiometabolic diseases are a major global health burden.
• Nuclear magnetic resonance (NMR) metabolomics offers precise assessment of metabolic individuality.

Purpose of the Study:

• To decode plasma metabolomic profiles using unsupervised learning.
• To gain new insights into the etiology of cardiometabolic diseases.

Main Methods:

• Unsupervised learning applied to plasma profiles of 118,001 UK Biobank participants.
• Phenome-wide and genome-wide association studies.
• Prospective cohort analyses and Mendelian randomization.

Main Results:

• Eleven distinct metabolic clusters identified, linked to 101 genetic loci and 445 phenotypes, predominantly cardiometabolic diseases.
• Novel metabolic signatures showed improved cardiometabolic risk prediction compared to traditional lipids.
• Triglyceride-rich lipoproteins outperformed apolipoprotein B and lipoprotein(a) in predicting ischemic heart disease, type 2 diabetes, and hypertension.

Conclusions:

• Metabolic signatures provide comprehensive, interpretable information for clinical translation.
• Lipid subpopulations play critical roles in cardiometabolic risks.
• A nuanced approach to blood lipid management is encouraged to balance disease risks.