Abstract 3352: Development and validation of a serum microRNA-based diagnostic model for early detection of multiple cancers

Abstract

Abstract Introduction: Detecting cancer early improves patients’ survival rate. We previously developed a serum 4-microRNA (miRNA) diagnostic model using lung cancer patient samples, which detected 12 cancer types with high accuracy (Zhang and Hu 2022 Cancers). In this study, we aimed to develop a new diagnostic model based on samples from ovarian cancer patients to evaluate whether model performance depends on the initial cancer type used. Patients and Methods: Our previous work assembled a serum miRNA microarray dataset from 4 individual datasets from GEO, including 3604 patients across 13 cancer types and 3932 non-cancer controls. In this study, we composed a training set of 191 ovarian cancer patients and 191 non-cancer controls matched by age and gender. Limma analysis was performed to rank differentially expressed miRNAs, and 10-fold cross-validation was used to determine the optimal number of top miRNAs for the final diagnostic model. Model performance was evaluated by area under the curve (AUC) of Receiver Operating Characteristic (ROC) curve analysis, and sensitivity and specificity, using samples not utilized in current or previous model development. Results: Using this training set, a 10-miRNA model was developed which showed an improved AUC 0.994 as compared to 0.973 from our previous model for ovarian cancer in the testing set. AUC was also substantially improved on sarcoma and breast cancers. With 99.1% specificity, the current model demonstrated at least 70% sensitivity for 10 of the 13 cancer types, comparable to the previous model with 99.3% specificity (Table 1). Conclusion: Our study showed that blood miRNA-based diagnostic models can be developed for detecting multiple cancers with high accuracy, regardless of initial cancer types used in model development. These results suggested that tumors of different cancers shed common types of miRNAs into the bloodstream, supporting the use of circulating miRNAs as a biomarker for multi-cancer early detection. ABLE 1. Diagnostic Model Performance Cancer Type N Current Model AUC Current Model Sensitivity Previous Model AUC Previous Model Sensitivity Ovarian 142 0.994 0.80 0.973 0.69 Lung 1358 0.998 0.93 0.999 0.99 Pancreatic 149 0.992 0.76 0.983 0.83 Prostate 40 0.998 0.95 0.996 0.92 Sarcoma 132 0.973 0.52 0.876 0.72 Esophageal 124 0.995 0.90 0.990 0.85 Gastric 150 0.998 0.93 0.999 1.00 Glioma 40 0.985 0.68 0.996 0.88 Liver 348 0.988 0.70 0.979 0.84 Biliary Tract 40 0.996 0.85 0.998 0.98 Bladder 392 0.996 0.95 0.998 0.98 Breast 135 0.946 0.10 0.378 0.00 Colorectal 155 0.99 0.83 0.955 0.86 Citation Format: Grace Zhou, Hai Hu. Development and validation of a serum microRNA-based diagnostic model for early detection of multiple cancers [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 3352.