Diagnostic potential for a serum miRNA neural network for detection of ovarian cancer

• Published in: eLife Vol. 6
• Main Authors: Elias, Kevin M, Fendler, Wojciech, Stawiski, Konrad, Fiascone, Stephen J, Vitonis, Allison F, Berkowitz, Ross S, Frendl, Gyorgy, Konstantinopoulos, Panagiotis, Crum, Christopher P, Kedzierska, Magdalena, Cramer, Daniel W, Chowdhury, Dipanjan
• Format: Journal Article
• Language: English
• Published: England eLife Sciences Publications Ltd 31.10.2017; eLife Sciences Publications, Ltd
• Subjects: Age; Aged; Biomarkers, Tumor - blood; Cancer Biology; Cancer therapies; Epidemiology; FDA approval; Female; Gene expression; Gynecology; Health care; Hospitals; Humans; Machine learning; Medical diagnosis; Medical screening; Metastases; MicroRNAs - blood; Middle Aged; miRNA; neural network; Neural networks; Neural Networks, Computer; next generation sequencing; Obstetrics; Oncology; Ovarian cancer; Ovarian Neoplasms - diagnosis; Ovarian Neoplasms - pathology; Pathology; Predictive Value of Tests; Prostate; Real-Time Polymerase Chain Reaction; Sensitivity and Specificity; serum; Serum - chemistry; Surgery; Tumors; Womens health; ovarian cancer; next generation sequencing; miRNA; cancer biology; machine learning; serum; human; neural network
• ISSN: 2050-084X; 2050-084X
• DOI: 10.7554/eLife.28932

Recent studies posit a role for non-coding RNAs in epithelial ovarian cancer (EOC). Combining small RNA sequencing from 179 human serum samples with a neural network analysis produced a miRNA algorithm for diagnosis of EOC (AUC 0.90; 95% CI: 0.81–0.99). The model significantly outperformed CA125 and functioned well regardless of patient age, histology, or stage. Among 454 patients with various diagnoses, the miRNA neural network had 100% specificity for ovarian cancer. After using 325 samples to adapt the neural network to qPCR measurements, the model was validated using 51 independent clinical samples, with a positive predictive value of 91.3% (95% CI: 73.3–97.6%) and negative predictive value of 78.6% (95% CI: 64.2–88.2%). Finally, biologic relevance was tested using in situ hybridization on 30 pre-metastatic lesions, showing intratumoral concentration of relevant miRNAs. These data suggest circulating miRNAs have potential to develop a non-invasive diagnostic test for ovarian cancer. Ovarian cancer is a major cause of cancer death among women. A woman’s survival often hinges on doctors detecting the tumor before it has spread beyond the ovary. Unfortunately, most women with ovarian cancer are not diagnosed until they have symptoms – such as pelvic pain, bloating, swelling of the abdomen or appetite loss. By then, the disease has usually spread and is difficult to treat. There is currently no reliable test to diagnose ovarian cancer before symptoms emerge. Some tests measure proteins in the blood or use ultrasound images to identify ovary tumors. These tests usually still identify the disease too late. Sometimes they produce “false positive” results, which may cause women without cancer to undergo unnecessary surgery. Many ovarian cancers have defects in small pieces of genetic information called microRNAs. These microRNAs impact the tumor in multiple ways, and cells release microRNAs into the blood. Testing a seemingly healthy women’s blood for the same pattern of altered microRNAs found in women with ovarian cancer might be one way to detect the disease earlier. Now, Elias et al. have identified a pattern of seven microRNAs in the blood that appears to predict ovarian cancer. In the experiments, a computer program searched for microRNA patterns in women with ovarian cancer. The program sifted through the microRNAs in blood from women with and without ovarian cancer. Over time, the computer program “learned” to identify a pattern of microRNAs found only in women with ovarian cancer. It then created a formula for identifying ovarian cancer based on seven of the microRNAs. Elias et al. then verified that the formula accurately detected ovarian cancer by testing it on blood samples from more women with and without cancer. They also found the seven microRNAs in tiny ovarian cancer tumors collected from women. This suggests the formula might be able to detect even the smallest tumors. More studies are needed to determine when this cancer-linked pattern first emerges and confirm that this ovarian cancer-detection formula works. If the test is validated, it might be used to screen women who are at high risk for ovarian cancer because of mutations in the BRCA1 and BRCA2 genes.