Enhancing ovarian cancer prognosis with an artificial intelligence-derived model: Multi-omics integration and therapeutic implications

You Wu; Kunyu Wang; Yan Song; Bin Li

doi:10.1016/j.tranon.2025.102439

Enhancing ovarian cancer prognosis with an artificial intelligence-derived model: Multi-omics integration and therapeutic implications

Transl Oncol. 2025 Jun 27:59:102439. doi: 10.1016/j.tranon.2025.102439. Online ahead of print.

Authors

You Wu¹, Kunyu Wang¹, Yan Song², Bin Li³

Affiliations

¹ Department of Gynecology Oncology, National Cancer Center/National Clinical Research Center for Cancer/Cancer Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, China.
² Department of Pathology, National Cancer Center/National Clinical Research Center for Cancer/Cancer Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, China. Electronic address: songyan@cicams.ac.cn.
³ Department of Gynecology Oncology, National Cancer Center/National Clinical Research Center for Cancer/Cancer Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, China. Electronic address: libin@cicams.ac.cn.

PMID: 40580871
DOI: 10.1016/j.tranon.2025.102439

Abstract

Background: Gynecological malignancies, particularly ovarian cancer, pose a formidable challenge to women's wellbeing, as evidenced by the global incidence and mortality rates, emphasizing the pressing need for advanced diagnostic and treatment modalities. The heterogeneity of ovarian cancer poses challenges for traditional therapeutic approaches, necessitating the exploration of novel, precision medicine techniques.

Methods: This study leveraged multi-dataset analysis to construct and validate an Artificial Intelligence-Derived Prognostic Index (AIDPI) for ovarian cancer. Transcriptome data from the TCGA, ICGC, and GEO databases were utilized, encompassing bulk and single-cell RNA sequencing. The AIDPI model was developed and refined using univariate Cox regression analysis and an ensemble of machine learning algorithms. Functional analysis, immunoprofiling, and the role of the MFAP4 gene were investigated to elucidate the biological mechanisms underlying the model.

Results: The AIDPI model demonstrated superior accuracy in predicting ovarian cancer prognosis compared to existing models. It correlated with clinical treatment outcomes, including chemotherapy responsiveness, and was integrated into a nomogram for improved prognostic stratification. Functional analysis revealed the influence of AIDPI genes on tumor immune infiltration and cell cycle regulation. Single-cell analysis exposed cell type-specific expression patterns, and the MFAP4 gene was identified as a potential therapeutic target due to its association with patient prognosis and modulation of cellular behavior. In clinical samples of ovarian cancer patients, MFAP4 is highly expressed in metastatic lesions and is associated with poor prognosis. In vitro and in vivo experiments, knockdown of MFAP4 reduces the metastasis of ovarian cancer cells.

Conclusion: The AIDPI model offers a highly accurate tool for ovarian cancer prognosis and treatment decision-making, underscored by the integration of multi-omics data and artificial intelligence. The model's performance and biological insights provide a foundation for advancing precision medicine in ovarian cancer. MFAP4's functionality and the influence of DNA methylation present opportunities for prospective research endeavors and potential therapeutic interventions.

Keywords: AIDPI; Immune microenvironment; MFAP4; Ovarian Cancer; Precision medicine; Prognostic model.