Development and validation of a proteo-metabolic panel for detection of asymptomatic ovarian cancer with minimal serum sample requirements: A multi-center prospective study.

Abstract

5546 Background: Early diagnosis is crucial for improving the prognosis of ovarian cancer (OC). However, most patients (pts) are diagnosed at advanced stages due to subtle symptoms and traditional biomarkers' limitations. We aimed to develop an serum panel using multi-omics data for cost-effective detection of asymptomatic OC (asym-OC). Methods: Participants were recruited from the Shanghai Ovarian Cancer and Family Care Project ( NCT06118307 ) involving five centers. A total of 843 individuals were included: 135 asym-OC pts and 708 non-OC individuals (290 pts with benign lesions and 418 healthy controls). Fasting serum samples (1 μL each) were analyzed using MALDI-TOF MS to generate proteo-metabolic (pro-met) data. For each sample, the original MS have ~43,900 data points from 100-13,000 Da. Preoperative data from three centers (N=680) were used to develop Light Gradient Boosting Machine (LGBM) models to identify key signals differentiating OC from non-OC. An independent external validation set (N = 163) was assembled by the other two centers. To develop a biologically interpretable and generalizable panel, the model was further refined to minimize biomarkers while maintaining efficacy. Validation of the panel was conducted using postoperative pro-met data from 42 asym-OC pts, transcriptomic data from 89 with asym-OC and 39 with benign lesions, supported by bioinformatic analyses. Results: Ten biomarkers, which are involved in coagulation, complement system, carcinogenesis, epithelial-mesenchymal transition, and the Warburg effect, were selected in the panel. The panel achieved an AUC of 0.90 (95% CI: 0.82-0.98) in the external validation set. The enhanced model that integrating the panel, age, BMI and HE4 achieved an AUC of 0.94 (95% CI: 0.90-0.99). In subgroup analyses, the enhanced model outperformed CA125, HE4, and ROMA, with AUCs of 0.96 (early-stage OC vs. non-OC) and 0.96 (OC vs. endometriosis) (See Table). Moreover, postoperative levels of the biomarkers in the panel approached those of the non-OC (p < 0.05). Transcriptomic data of the tissues corresponded to the pathophysiological alterations associated with OC development or progression. Conclusions: This study introduces a novel, non-invasive, and cost-effective serum panel that demonstrates high sensitivity and specificity for detecting asym-OC, offering a promising tool for early diagnosis of the disease. Early-stage OC vs. Non-OC (N=58/708) OC vs. Endometriosis (N=135/80) Enhanced model: 10-biomarker Pro-Met panel + BMI +age + HE4 0.96 [0.94-0.99] 0.93 [0.90-0.97] CA125 0.87 [0.82-0.93] 0.77 [0.71-0.83] HE4 premenopausal 0.73 [0.63-0.83] 0.91 [0.80-0.94] postmenopausal 0.67 [0.58-0.74] 0.88 [0.76-0.86] ROMA premenopausal 0.75 [0.60-0.81] 0.91 [0.80-0.93] postmenopausal 0.89 [0.67-0.84] 0.87 [0.53-1.00]