ABSTRACT

Background and Aim: Early pregnancy diagnosis is essential for improving reproductive efficiency and reducing non-productive days in commercial swine production. Conventional ultrasonography generally provides reliable pregnancy confirmation only 23–25 days after insemination, potentially missing the optimal opportunity to rebreed non-pregnant sows during the estrous cycle. Vaginal secretions represent a readily accessible and minimally invasive biological matrix that may reflect early pregnancy-associated metabolic alterations. This study aimed to identify metabolite biomarkers in vaginal secretions for early pregnancy diagnosis in sows using integrated untargeted metabolomics and weighted co-expression network analysis (WGCNA).

Materials and Methods: Vaginal secretion samples were collected from healthy Landrace × Large White crossbred gilts at 18 days post insemination. Pregnancy status was confirmed by ultrasonography at 23–25 days post insemination, resulting in 15 pregnant and 10 non-pregnant animals. Untargeted metabolomic profiling was performed using liquid chromatography-mass spectrometry. Differential metabolites were identified using variable importance in projection ≥1 and adjusted p < 0.05, followed by WGCNA to identify pregnancy-associated metabolite modules and hub metabolites. Kyoto Encyclopedia of Genes and Genomes pathway enrichment analysis and receiver operating characteristic (ROC) curve analysis were performed to evaluate biological relevance and diagnostic performance.

Results: A total of 3,249 metabolic features were detected, of which 534 metabolites differed significantly between pregnant and non-pregnant sows. Differential metabolites were primarily enriched in nucleotide metabolism and lipid metabolism-related pathways. WGCNA identified three metabolite modules significantly associated with pregnancy status (p < 0.01). Four biologically relevant hub metabolites, progesterone, prostaglandin B1, lithocholic acid, and taurolithocholic acid, were identified as candidate biomarkers. Individually, all four metabolites achieved ROC area under the curve (AUC) values exceeding 0.80, while a combined four-metabolite panel achieved an AUC of 0.89, with 80% sensitivity and 100% specificity for distinguishing pregnant from non-pregnant sows. These findings indicate that integrating metabolomics with network analysis improves biomarker discovery beyond conventional differential metabolite screening.

Conclusion: Integrated untargeted metabolomics and WGCNA identified four promising vaginal secretion biomarkers for early pregnancy diagnosis in sows as early as 18 days post insemination. The combined biomarker panel demonstrated excellent diagnostic performance and supports the potential development of a rapid, non-invasive diagnostic assay for commercial swine production. Validation in larger, independent populations and development of practical on-farm detection platforms are warranted.

Keywords: biomarkers, early pregnancy diagnosis, liquid chromatography-mass spectrometry, metabolomics, sows, vaginal secretions, weighted co-expression network analysis.