Serum Multiomics Profiling of Infertile Women Stratified by Etiology Reveals Distinct Metabolic, Lipidomic, and Exposomic Signatures

Abstract Female infertility is evaluated primarily by hormonal assays and anatomical imaging, and its molecular landscape remains largely uncharacterized. We enrolled 247 women of reproductive age: 57 nonpregnant controls (NP), 59 women with normal pregnancies (Preg), and 131 infertile patients stratified into polyendocrine metabolic ovarian syndrome-related (PMOS, n = 49), organic/anatomical/inflammatory (OAII, n = 26), other endocrine/autoimmune (OEAI, n = 8), and unexplained (UEI, n = 48) infertility. Fasting serum underwent untargeted metabolomic, lipidomic, and exposomic profiling; classifiers were built with nested cross-validation and integrated via DIABLO, and exposure mixtures were assessed by BKMR. The integrated classifier separated PMOS (AUC 0.892) and UEI (AUC 0.993) from NP, whereas OAII was less separable (AUC 0.729, permutation p = 0.14); lipidomics alone outperformed the integrated model for PMOS and OAII (both DeLong p < 0.001). PMOS was marked by long-chain lipid upregulation, OAII by pan-lipidomic upregulation, and UEI by global lipid downregulation; etiology-stratified sensitivity analyses indicated that OAII heterogeneity diluted, rather than drove, its pooled signal. The pregnant reference group showed coordinate lipidomic and metabolomic remodeling that opposed the infertile acylcarnitine signature. Octabenzone (PIP > 0.98) and propylparaben (PIP 0.75–1.00) emerged as key exposures. Lipidomics emerged as the most informative layer, supporting precision molecular phenotyping of female infertility.