Integrative analysis of tumor-educated platelets for stage-specific diagnosis, prognosis, and therapy in ovarian cancer

BACKGROUND: Ovarian cancer (OC) remains one of the most lethal gynecological cancers worldwide. Despite advances in diagnosis, OC is mostly detected at late stages due to undefined symptoms. Therefore, identifying feasible, reliable, non-invasive biomarkers for early detection and disease stratification of OC is crucial. Tumor-educated platelets (TEPs) have emerged as a promising source for liquid biopsy, harboring oncogenic mRNA signatures that reflect the tumor microenvironment. In this study, we investigated TEP-mRNAs to identify stage-specific biomarkers for OC diagnosis and prognosis, while uncovering disease progression mechanisms and potential therapeutic targets. METHODS: An integrative framework that combines machine learning (ML) techniques with differential expression analysis was used to identify candidate biomarker genes (CBGs) across OC stages. Random forests were used to assess the diagnostic accuracy of CBGs, and survival analysis was used to evaluate their prognostic significance. In silico drug screening was performed to prioritize potential anticancer drugs identified via a drug-gene network. Finally, pathway enrichment analysis was performed to elucidate the progression mechanisms specific to each stage. Multiple datasets have been utilized for analysis and validation. RESULTS: The proposed integrative analysis revealed 50 CBGs across the four OC stages. Predictive models have been trained utilizing the CBGs expression profiles and achieved F1-scores of 95%, 95%, 80%, and 93% for stages 1, 2, 3, and 4, respectively. Focusing on CBGs with prognostic relevance, survival analysis identified seven candidate genes associated with patient outcomes: PIK3AP1 and IARS2 (stage 1); CTSW, QSOX1, and CASP1 (stage 2); GZMB (stage 3); and CA1 (stage 4). In silico drug screening and drug-gene network analysis prioritized 21 FDA-approved drugs for further evaluation. Pathway enrichment analysis revealed 22 stage-specific pathways (SSPs), highlighting key molecular changes in OC, including kinesins and collagen chain trimerization for early stages. CONCLUSION: The proposed framework integrates ML and bioinformatics techniques for stage-specific biomarker discovery in OC using TEP mRNAs. The candidate biomarkers exhibit prognostic and diagnostic potential. Beyond early detection, the framework supports the potential clinical utility of TEP mRNAs for risk stratification and patient classification. Furthermore, the identified drug candidates provide a basis for further investigation toward more targeted treatment approaches. Overall, this study highlights the promising role of TEP mRNAs as a non-invasive tool for early diagnosis and prognosis.