Integrating machine learning and metabolomics to uncover new biomarkers for predicting pesticide exposure among patients with kidney function decline.

Although pesticide application is indispensable for agricultural productivity, improper use can pose significant health risks, particularly for vulnerable populations. This study investigated the effects of pesticide exposure on metabolic pathways and disease progression among patients with chronic kidney disease (CKD). This longitudinal study enrolled 89 CKD patients. A total of 71 pesticides, including 9 carbamate pesticides, were detected in the urine samples. Our findings indicated that higher concentrations of carbamates, possibly from the diet, may significant be associated with oxidative stress, amino acid metabolism, and mitochondrial energy metabolism in CKD patients. Integrating machine learning approach identified l-glutamine (L-Glu), 3-chlorotyrosine, and N2,N2-dimethylguanosine as potential biomarkers of pesticide exposure, with an area under the curve of >0.903 based on machine learning- a receiver operating characteristic analysis. Both CKD and pesticide exposure were associated with abnormalities in amino acid and energy metabolism. Key metabolites such as L-cysteine, Acetyl-CoA, L-Glu, and L-histidine were identified as endogenous markers capable of predicting changes in both renal dysfunction and pesticide exposure among CKD patients. Detecting exposure-related metabolic alterations through metabolomics enables early identification and aids in the understanding and potential prevention of kidney disease progression. This study further can explore the clinical applicability and improve predictive value of these biomarkers.