| ObjectivesThis study is aimed to investigate the availability of tissue specificity index(TSI)to analyze circulating microRNA(miRNA)datasets using high-throughput reverse transcriptional quantitative polymerase chain reaction(RT-qPCR),establish a novel global index of quality evaluation(GIQE),and compare the pattern of GIQE on reference selection with geNorm and NormFinder algorithms,which could provide a novel strategy for quality control and reference selection of high-throughput RT-qPCR circulating miRNA datasets.MethodsDownloading high-throughput RT-qPCR circulating miRNA datasets,processing pre-analytic data and computing their GIQE stability scores ranged from 0 to 1were performed using series of packages including GEOquery,miRbase.db,read.qPCR and NormqPCR package,etc in R Statistic and Computing Console.The ability of GIQE score on distinguishing co-expressed miRNAs from those partially co-expressed circulating miRNAs in each dataset was evaluated using the Kruskal-Wallis test,receiver operating characteristic(ROC)curve and multivariate linear regression analysis.Correlations among average Ct values of GIQE-,geNorm-and NormFinder-selected reference candidates were evaluated.Results1 Basic characterization of high-throughput RT-qPCR circulating miRNA was datasets(1.2.1).Downloading 11 circulating miRNA datasets including series of sample types including plasma,serum,exosome and peripheral blood mononuclear cell(PBMC)in case-control study.2 Testing GIQE score of circulating miRNA stability in sample of each dataset(1.2.3).The GIQE score ordered from high to low stability subsequently are0.089[0.061,0.163]for GSE86226,0.188[0.145,0.341]for GSE79922,0.205[0.146,0.528]for GSE79960,0.279[0.190,0.403]for GSE70318,0.440[0.244,0.856]for GSE65708,0.490[0.193,0.855]for GSE68314,0.561[0.270,0.916]for GSE47125,0.564[0.268,0.857]for GSE75389,0.592[0.260,0.904]for GSE96621,0.598[0.344,0.884]for GSE75391,and 0.760[0.497,0.938]for GSE70080,respectively.Non-parameteric testing showed that the miRNA GIQE scores had significant difference among 11 datasets(?~2=929,P<0.001).The GIQE scores of both GSE86226 with highest stability and GSE70080with lowest stability significantly exhibited lower and higher than those of the other datasets(GSE79922:Z=-6.64,P<0.001 and Z=17.22,P<0.001;GSE79960:Z=-8.29,P<0.001 and Z=14.89,P<0.001;GSE70318:Z=-7.36,P<0.001 and Z=13.44,P<0.001;GSE65708:Z=-16.79,P<0.001 and Z=9.38,P<0.001;GSE68314:Z=-17.54,P<0.001 and Z=11.73,P<0.001;GSE47125:Z=-18.72,P<0.001 and Z=6.24,P<0.001;GSE75389:Z=-16.01,P<0.001 and Z=4.87,P<0.001;GSE96621:Z=-18.57,P<0.001 and Z=6.76,P<0.001;GSE75391:Z=-17.78,P<0.001 and Z=4.08,P<0.001).Multivariate linear regression analysis showed that the median values of GIQE score in each dataset were associated with the number of both total miRNAs(?=0.292±0.061,t=4.818,P=0.003)and co-expressed miRNAs(?=-0.224±0.053,t=-4.144,P=0.006)(F=20.38,P=0.001).3 ROC analysis as shown in 1.2.4 showed that the AUC values were higher than 0.900,the optimal cut-off values ranged from 0.216 to 0.486 with the specificity of 0.851 to 0.963 and the sensitivity of 0.888 to 0.977.4 Correlation analysis of average Ct values between geNorm-and NormFinder-selected reference candidates(1.2.5).The average Ct values with only 10%geNorm-selected candidates for GSE68314 were not associated with those selected by NormFinder(r=0.798,P=0.127),whereas the values of 100%,50%,25%or 10%geNorm-selected candidates in all the other datasets were individually associated with the corresponding NormFinder-selected candidates and the coefficients ranged from 0.8 to 1.0,which exhibited strong correlation between them.5 Correlation analysis of the average Ct values in samples each dataset between GIQE-and either geNorm-or NormFinder-selected candidates(1.2.6).The values of the GIQE-selected candidates were associated with those of either geNorm-or NormFinder-selected candidates,respectively,along with the increased ratio of the selected candidates to 50%(GSE79922:r=0.948[0.791,0.988],P<0.001(geNorm)and r=0.960[0.837,0.991],P<0.001(NormFinder);GSE70318:r=0.985[0.976,0.990],P<0.001(geNorm)and r=0.983[0.973,0.989],P<0.001(NormFinder);GSE47125:r=0.984[0.950,0.995],P<0.001(geNorm)and r=0.975[0.923,0.992],P<0.001(NormFinder);GSE75391:r=0.964[0.888,0.989],P<0.001(geNorm)and r=0.911[0.737,0.972],P=0.006(NormFinder);GSE79960:r=0.860[0.503,0.966],P=0.02(geNorm)and r=0.929[0.720,0.983],P=0.005(NormFinder);GSE68314:r=0.905[0.477,0.986],P=0.003(geNorm)and r=0.865[0.321,0.980],P=0.005(NormFinder);GSE96621:r=0.942[0.830,0.981],P<0.001(geNorm)and r=0.849[0.595,0.949],P=0.002(NormFinder);GSE65708:r=0.917[0.753,0.974],P=0.005(geNorm)and r=0.932[0.794,0.979],P=0.004(NormFinder);GSE75389:r=0.886[0.264,0.987],P=0.019(geNorm)and r=0.843[0.099,0.982],P=0.035(NormFinder)).Finally,the percentage of the common candidates ranked between 0 to 50%among three algorithms was 8%-90%for GIQE,50%-90%for geNorm,61%-100%for NormFinder,respectively.Conclusions1 The TSI-transformed GIQE can estimate the quality of all miRNA raw data independently;2 GIQE enables quantitative assessment of miRNA expression pattern classification;3 The ratio of more than 50%of GIQE reference candidates can achieve the effect of geNorm and NormFinder;4 GIQE provides a novel quantifiable strategy for quality control and reference selection of high-throughput RT-qPCR circulating miRNA dataset.Figure12;Table27;Reference 111... |
PDF Full Text Request