Parallel multi-omics profiling of endometrial cancer tissue and serum with development of a combined serum biomarker model

The aim of this study was to integrate proteomic and metabolomic datasets derived from tissue and serum samples of patients with endometrial cancer (EC) to comprehensively characterize the molecular landscape associated with the disease. A total of 60 patients with EC and 32 healthy controls were included. Tissue and serum samples were analyzed using data-independent acquisition proteomics together with metabolomics platforms. Molecular characteristics were investigated through differential analyses followed by pathway enrichment evaluation. In addition, machine learning approaches were applied to identify serum-derived biomarkers and to develop predictive diagnostic models. A total of 8,542 proteins and 576 metabolites were identified in tissue samples, whereas 465 proteins and 405 metabolites were identified in serum samples. Tissue-level analysis demonstrated predominant enrichment in oncogenic signaling pathways, including the RAS-RAF-MAPK cascade, along with metabolic pathways associated with amino acid and nucleotide metabolism. In contrast, serum profiles reflected broader systemic alterations, particularly involving immune-related processes, platelet activation, complement cascade pathways, and metabolism of branched-chain amino acids. Integration of serum multi-omics data identified CFHR3, LAMP2, L-valine, and L-isoleucine as candidate biomarkers. The multi-omics combined predictive model achieved an average area under the curve of 0.98, demonstrating substantially improved performance compared with models based on individual proteins or metabolites. Distinct molecular characteristics were observed between EC tissue and serum. The serum-based multi-omics predictive model demonstrated strong diagnostic performance and identified a candidate biomarker panel with potential utility for non-invasive detection of EC.