Metabolomic machine learning predictor for arsenic-associated hypertension risk in male workers

Abstract

Arsenic (As)-induced hypertension is a significant public health concern, highlighting the need for early risk prediction. This study aimed to develop a predictive model for occupational As exposure and hypertension using metabolomics and machine learning. A total of 365 male smelting workers from southern regions were selected. Forty workers from high and low urinary arsenic (U-As) exposure groups were chosen for non-targeted metabolomics analysis. Univariate analysis revealed that U-As is a risk factor for blood pressure and hypertension (P < 0.05). Restricted cubic spline (RCS) analysis showed that both systolic and diastolic blood pressure, as well as hypertension risks, increased with U-As, with a threshold at 32 µg/L. Of 1145 metabolites, 383 differentially expressed metabolites (382 upregulated, 1 downregulated) were identified. Least absolute shrinkage and selection operator (LASSO) regression was used to construct a predictive model for occupational hypertension, with N-hexosyl leucine, myristic acid, gamma-glutamylvaline, and pregnanediol disulfate as predictors. The area under the curve (AUC) of the receiver operating characteristic (ROC) for the predictive model was 0.917, indicating strong predictability and accuracy. This model, based on metabolomics and machine learning, provides an effective tool for early identification and intervention for occupational populations at high risk of hypertension due to As exposure.