This study aimed to evaluate the role of circulating metabolites in enhancing the performance of the risk of ovarian malignancy algorithm (ROMA) for predicting the risk of ovarian cancer in women with ovarian cysts. The first batch of sera from 101 people with ovarian cancer (both serous and non serious) and 134 people with benign pelvic benign pelvic masses (BPM) tumors was subjected to metabolomic profiling. To tell early-stage ovarian cancer from BPM using a deep learning model containing 7 cancer-related metabolites: diacetylspermine, diacetylspermidine, N-(3-acetamidopropyl)pyrrolidin-2-1, N-acetylneuraminate, N-acetyl-mannosamine, N-acetyl-lactosamine, and hydroxy. 

The performance of the metabolite panel was assessed in a separate collection of sera from 118 cases of ovarian cancer and 56 patients with BPM. In addition, the panel’s efforts to enhance ROMA’s functionality were also evaluated. In the validation set, a 7-marker metabolite panel (7MetP) created for early-stage ovarian cancer had an AUC of 0.86 [95% CI: 0.76-0.95]. For early-stage ovarian cancer in the test set, the 7MetP+ROMA model had an AUC of 0.93 (95% CI: 0.84-0.98), which was superior to ROMA alone [0.91 (95% CI: 0.84-0.98); likelihood ratio test P: 0.03]). 

When applied to all specimens, the 7MetP+ROMA model had a better positive predictive value for detecting ovarian cancer at an early stage than ROMA alone (0.68 vs. 0.52; one-sided P<0.001) and a higher specificity for doing so (0.89 vs. 0.78; one-sided P<0.001). To better enhance clinical decision-making, researchers created a blood-based metabolite panel that displays independent predictive capacity and supports ROMA in identifying ovarian cancer in its earliest stages.