Construct of qualitative diagnostic biomarkers specific for glioma by pairing serum microRNAs

Abstract

Background Serum microRNAs (miRNAs) are promising non-invasive biomarkers for diagnosing glioma. However, most reported predictive models are constructed without a large enough sample size, and quantitative expression levels of their constituent serum miRNAs are susceptible to batch effects, decreasing their clinical applicability. Methods We propose a general method for detecting qualitative serum predictive biomarkers using a large cohort of miRNA-profiled serum samples (n = 15,460) based on the within-sample relative expression orderings of miRNAs. Results Two panels of miRNA pairs (miRPairs) were developed. The first was composed of five serum miRPairs (5-miRPairs), reaching 100% diagnostic accuracy in three validation sets for distinguishing glioma and non-cancer controls (n = 436: glioma = 236, non-cancers = 200). An additional validation set without glioma samples (non-cancers = 2611) showed a predictive accuracy of 95.9%. The second panel included 32 serum miRPairs (32-miRPairs), reaching 100% diagnostic performance in training set on specifically discriminating glioma from other cancer types (sensitivity = 100%, specificity = 100%, accuracy = 100%), which was reproducible in five validation datasets (n = 3387: glioma = 236, non-glioma cancers = 3151, sensitivity> 97.9%, specificity> 99.5%, accuracy> 95.7%). In other brain diseases, the 5-miRPairs classified all non-neoplastic samples as non-cancer, including stroke (n = 165), Alzheimer’s disease (n = 973), and healthy samples (n = 1820), and all neoplastic samples as cancer, including meningioma (n = 16), and primary central nervous system lymphoma samples (n = 39). The 32-miRPairs predicted 82.2 and 92.3% of the two kinds of neoplastic samples as positive, respectively. Based on the Human miRNA tissue atlas database, the glioma-specific 32-miRPairs were significantly enriched in the spinal cord (p = 0.013) and brain (p = 0.015). Conclusions The identified 5-miRPairs and 32-miRPairs provide potential population screening and cancer-specific biomarkers for glioma clinical practice.

Integrated analysis of diverse cancer types reveals a breast cancer-specific serum miRNA biomarker through relative expression orderings analysis

“…To ensure the reliability of the data, we removed outlier samples from each phenotype in each data set [ 32 ]. Briefly, we first calculated the correlation coefficients between the expression levels of miRNAs of any two samples.…”

“…The REO of this miRPair within this sample is either E miRNAa > E miRNAb or E miRNAa ≤ E miRNAb . If the REO distribution in two groups of samples is significantly different, then this REO can be used to predict the group to which an unknown sample belongs [ 32 ].…”

MicroRNA-nanoparticles against cancer: Opportunities and challenges for personalized medicine

“…The second panel includes 32 serum miRPairs, reaching 100% diagnostic performance in specifically discriminating glioma from other cancer types. 20 Tissue- and plasma extracellular vesicles-derived miR-99a-5p have been reported as potential biomarkers for predicting early head and neck squamous cell carcinoma. 21 Tumor tissue exhibited lower miR-99a-5p levels than para-tumor tissue while plasma extracellular vesicles harbored more miR-99a-5p compared with benign disease group.…”