Characterization of the Human Exposome and Risk of Reproductive Hormone Disruption by Integrating GC-HRMS Nontarget Analysis, Molecular Networking, and Knowledge Transfer Learning

Chemical characterization of the human exposome has largely overlooked nonpolar substances analyzed by gas chromatography-high-resolution mass spectrometry (GC-HRMS). In this study, nontarget analysis by GC-HRMS was applied to screen nonpolar exogenous contaminants in the whole blood ( n = 155) of healthy adults in the general population of China. With the assistance of molecular networking, 194 molecular features were annotated after excluding 55 features by blank filtering: 8 annotated at confidence level 1, 173 at level 2, and 13 at level 3. Among these, 93 were noncyclic alkanes or alkane derivatives not previously reported in exposome studies. These alkanes were accompanied by aromatic compounds (58), halogenated compounds (21), heterocyclic compounds (16), and cycloalkanes (6). Acknowledging the small dataset and environmentally relevant low exposure levels, a transfer learning model was implemented to predict the cumulative hormone-disrupting potential of the newly identified chemicals and of previously detected perfluoroalkyl substances (PFAS). PFAS primarily contribute to disruption of follicle-stimulating hormone, while GC-HRMS-detected aromatics make the greatest cumulative contribution to luteinizing hormone and estradiol. Unexpectedly, alkane derivatives contribute a 1.7-fold higher reproductive risk to urban/suburban residents than rural residents. The results will require confirmation in further work but underscore the importance of GC-HRMS for chemical exposomics by revealing understudied aromatic and alkane derivatives as risk factors.