A four-metabolite urine and plasma panel separated cancer cases from controls across eight cancer types, and a model that never used outcome labels still reached an AUC of 0.925 against 0.991 for supervised learning.
In 24,638 UK Biobank participants with prediabetes, LASSO and elastic net selection of nuclear magnetic resonance metabolomic markers yielded 16 biomarkers, and k-means clustering in the training set (12,322 participants) identified low-, intermediate-, and high-metabolic-risk subtypes that showed progressively higher risks of type 2 diabetes, cardiovascular disease, and chronic kidney disease in the validation set (12,316 participants), with diet-quality associations differing across subtypes and Mendelian randomization suggesting potential causal links for several metabolomic biomarkers.
Integrating plasma metabolomics, Mendelian randomization, and machine learning, this study identified phenylalanine as causally associated with pancreatic cancer among 55 plasma metabolites (IVW OR = 1.641, 95% CI 1.052–2.562, p = 0.029), derived eight related differentially expressed genes, built a random forest diagnostic model from 113 combinations of 12 algorithms (training AUC 0.994; validation AUCs 0.918, 0.983, 0.923), used SHAP to rank SLC6A14 as the top feature, and combined single-cell sequencing, simulated gene knockout, molecular docking, and molecular dynamics to suggest genistein binds SLC6A14 stably, with RT-qPCR confirming high expression of the five model genes in a BxPC-3 versus HPDE6-C7 cell pair.
You’re all caught up