Prioritizing key synergistic circulating microRNAs for the early diagnosis of biliary tract cancer

Abstract

Biliary tract cancer (BTC) is a highly aggressive malignant tumor. Serum microRNAs (ser-miRNAs) serve as noninvasive biomarkers to identify high risk individuals, thereby facilitating the design of precision therapies. The study is to prioritize key synergistic ser-miRNAs for the diagnosis of early BTC. Sampling technology, significant analysis of microarrays, Pearson Correlation Coefficients, t-test, decision tree, and entropy weight were integrated to develop a global optimization algorithm of decision forest. The source code is available at https://github.com/SuFei-lab/GOADF.git. Four key synergistic ser-miRNAs were prioritized and the synergistic classification performance was better than the single miRNA’ s. In the internal feature evaluation dataset, the area under the receiver operating characteristic curve (AUC) for each single miRNA was 0.8413 (hsa-let-7c-5p), 0.7143 (hsa-miR-16-5p), 0.8571 (hsa-miR-17-5p), and 0.9365 (hsa-miR-26a-5p), respectively, whereas the synergistic AUC value increased to 1.0000. In the internal test dataset, the single AUC was 0.6500, 0.5125, 0.6750, and 0.7500, whereas the synergistic AUC increased to 0.8375. In the independent test dataset, the single AUC was 0.7280, 0.8313, 0.8957, and 0.8303, and the synergistic AUC was 0.9110 for discriminating between BTC patients and healthy controls. The AUC for discriminating BTC from pancreatic cancer was 0.9000. Hsa-miR-26a-5p was a predictor of prognosis, patients with high expression had shorter survival than those with low expression. In conclusion, hsa-let-7c-5p, hsa-miR-16-5p, hsa-miR-17-5p, and hsa-miR-26a-5p may act as key synergistic biomarkers and provide important molecular mechanisms that contribute to pathogenesis of BTC.