Title:Meta-analysis of Microarrays: Detection of MiR-21 Expression as a Biomarker in Early Diagnosis of Non-small Cell Lung Cancer
Runing rubric:Analysis of miR-21 as diagnostic marker in NSCLC
A sum of four microarray datasets were retrieved in this meta-analysis.
MiR-21 look showed no difference between NSCLC patients and healthy people.
Significant heterogeneousness was observed.
Aim:Non-small cell lung malignant neoplastic disease ( NSCLC ) is one of the most common human malignances, of which patients are frequently diagnosed at advanced phase. MicroRNA-21 ( miR-21 ) , a possible biomarker in diagnosing of assorted human tumours, has been shown overexpressed in NSCLC. Thus it is necessary to execute a meta-analysis aimed at measuring the effectivity of miR-21 as a biomarker in early diagnosing of NSCLC.
Methods:Relevant microarray datasets about NSCLC-related miRNA were collected from NCBI Gene Expression Omnibus ( GEO ) database and EBI ArrayExpress database up to February 2014. Quality control was conducted by utilizing limma bundle and ExiMiR bundle in R. Meta bundle was used for the analysis of the information. Both random- and fixed-effects theoretical accounts were used to unite informations. Consequence sizes are presented as standardised mean difference ( SMD ) with 95 % assurance intervals ( CI ) . Heterogeneity was assessed utilizing the Chi-square trial of Q and the I2 statistic with significance set at P & A ; lt ; 0.0.05 or I2 & A ; gt ; 50 % . Sensitivity analysis was besides conducted to measure the stableness of the pooled consequences.
Consequences:A sum of four datasets ( GSE24704, GSE17681, GSE27486, and GSE40738 ) of miRNA in patients’ blood cells were selected including 153 NSCLC patients and 109 healthy people. The pooled consequences generated by a random-effect theoretical account revealed that no important difference was observed between instance group and control group ( SMD=0.58, 95 % CI: -0.04-1.19 ) . But important heterogeneousness was observed (P=0.0032, I2=78.2 % ) . Sensitivity analysis indicated that the consequences of present meta-analysis were unstable.
Decisions:It was concluded that miR-21 may has no relationship with the happening of NSCLC and it was uneffective to observe miR-21 look in early diagnosing of NSCLC.
Keywords:Non-small cell lung malignant neoplastic disease ; MiR-21 ; Microarray datasets ; Meta-analysis
Lung malignant neoplastic disease is the taking cause of cancer-related decease in the universe [ 1 ] . Non-small cell lung malignant neoplastic disease ( NSCLC ) , which makes up 85 % of all lung malignant neoplastic disease instances, is one of the most common human malignances [ 2 ] . Patients of NSCLC are frequently diagnosed at advanced phase or as metastatic disease [ 3 ] . These patients ( about history for 70 % ) ever lost the best period of operation because malignant neoplastic disease cells have already spread at the clip of diagnosing [ 4 ] .
MicroRNA ( miRNA ) is a little non-coding RNA molecular ( incorporating about 19-25 bases ) found in workss, animate beings, and some viruses, which has been found been related to malignant neoplastic disease development [ 5, 6 ] . It has been reported that degrees of miRNAs in serum are stable, consistent, and consistent among persons of the same species [ 7 ] . Chenet Al.concluded that serum miRNAs can function as possible biomarkers for the sensing of assorted malignant neoplastic diseases and other diseases [ 8 ] . MicroRNA-21 ( miR-21 ) , an oncogenic miRNA, has been shown to be overexpressed in assorted human tumours [ 9, 10 ] . The overexpression of miR-21 promotes tumorigenesis through suppression of negative regulators of Ras/MEK/ERK tract and suppression of programmed cell death [ 11 ] . Besides, Selaruet Al.demonstrated that miR-21 showed 95 % sensitive and 100 % specific in separating between cholangiocarcinomas and normal tissues [ 12 ] . And in diagnosing of lung malignant neoplastic disease, sensing of miR-21 look resulted in 69.66 % sensitiveness and 100.00 % specificity [ 13 ] . In NSCLC patients, the degree of serum miR-21 was important increased as compared with that in healthy people [ 14 ] . Thus it can be indicated that serum miR-21 look might be utile as serum diagnostic marker for supervising NSCLC development.
In present reappraisal, several cistron look profiling datasets about miR-21 look in NSCLC patients’ blood were collected from the online databases up to February 2014. Then a meta-analysis was performed to measure the suitableness of miR-21 as the biomarker in early diagnosing of NSCLC.
Materials and Methods
Microarray informations sets about NSCLC-related serum miRNA were collected from NCBI ( National Center of Biotechnology Information ) Gene Expression Omnibus ( GEO, hypertext transfer protocol: //www.ncbi.nlm.nih.gov/geo/ ) database and EBI ( The European Bioinformatics Institute ) ArrayExpress ( hypertext transfer protocol: //www.ebi.ac.uk/arrayexpress/ ) database up to February 2014. The hunt was based on the combination of following points: non-small cell lung malignant neoplastic disease ( or NSCLC ) , serum, blood, non-coding RNA profiling.
Datasets were included if they met the undermentioned standards: ( 1 ) Both NSCLC patients and healthy people were included in each dataset, and every group contained more than two samples ; ( 2 ) The look profiling informations of miRNA from instance group and control group were provided or calculated ; ( 3 ) The being weregay sapiens.
Quality control and informations extraction
Two research workers independently extracted the informations from all eligible datasets harmonizing to the standards listed supra. Any dissensions were resolved by treatment with a 3rd individual. Quality control, incorporating background rectification and standardization processing, was conducted by utilizing limma bundle [ 15 ] and ExiMiR bundle [ 16 ] in R. If multiple investigations were mapped to a individual miRNA, mean value of investigations was used as look value of miRNA. Therefore, look values of miR-21 in both instance groups and control groups were extracted. Means and standard divergences of these values were calculated. Sample sizes were counted for each dataset.
Meta-analysis was performed by utilizing meta bundle in R [ 17 ] . The consequence size of uninterrupted results was measured as standard mean difference ( SMD ) with 95 % assurance interval ( CI ) . Both random- and fixed-effects theoretical accounts were used to unite informations. Heterogeneity of consequence sizes across surveies was assessed by utilizing Chi-square trial of Q [ 18 ] and I2statistic [ 19 ] .Phosphorusvalue & A ; lt ; 0.05 or I2value & A ; gt ; 50 % was considered to be heterogenous, and so the random-effect theoretical account was selected. Pooled SMD and 95 % CI were calculated. OtherwisePhosphorus& A ; gt ; 0.05 or I2& A ; lt ; 50 % were homogenous, the fixed-effect theoretical account was chosen.
Sensitivity analysis was performed to measure the chief factors affected meta-analysis. The differences of pooled SMD and 95 % CI between fixed-effect theoretical account ( Mental-Haenszel method [ 20 ] ) and random-effect theoretical account ( Dersimonian-Laird method ) were analyzed. If same decision were obtained under two theoretical accounts, the consequences of meta-analysis were stable and dependable, otherwise the consequences were unstable.
Features of the selected datasets
In entire, four cistron look profiling datasets of miRNA in patients’ blood cells were selected. They were as the follows: GSE24709 ( Germany ) , GSE17681 ( Germany ) [ 21 ] , GSE27486 ( USA ) [ 22 ] and GSE40738 ( USA ) . MiRNA of four datasets wholly included 153 NSCLC patients and 109 healthy people. Blood from GSE24709 and GSE17681 were the peripheral blood, while in datasets GSE27486 and GSE40738, the blood used in analyses was the whole blood.
Analysis of heterogeneousness
Harmonizing to heterogeneousness trial, important difference was observed (Phosphorus=0.0032, I2=78.2 % , Fig. 1 ) . Therefore random-effect theoretical account was selected to cipher the pooled SMD and 95 % CI. No important difference was observed between instance group and control group ( SMD=0.58, 95 % CI: -0.04-1.19 ) .
Sensitivity analysis was performed for the four cistron look profiling datasets. The meta-analysis of miR-21 from fixed-effect theoretical account indicated that there were important difference between instance group and control group ( SMD=0.35, 95 % CI: 0.10-0.61, Fig. 2 ) . Meta-analysis from random-effect theoretical account revealed no important difference between the two groups ( SMD=0.58, 95 % CI: -0.04-1.19, Fig. 1 ) . Different decisions between two consequence theoretical accounts indicated that the consequences of present meta-analysis were unstable. And more datasets and samples should be included in farther meta-analysis to increase the stableness and dependability of the consequences.
In present meta-analysis, a sum of four cistron look profiling datasets were selected, incorporating 153 NSCLC patients and 109 healthy people. The pooled SMD revealed that there were no important differences between NSCLC group and control group. Besides, statistically important heterogeneousness and unstable consequences was observed by meta-analysis. Thus farther analysis should be performed with more datasets and samples to cut down instability and prejudice of the consequences.
Recently, miR-21 served as a biomarker for diagnosing of assorted malignant neoplastic diseases has been reported a batch. In pancreatic malignant neoplastic disease, plasma miRNAs including miR-21 demonstrated 64 % sensitiveness and 89 % specificity with the analyses of plasma degrees [ 23 ] . Similarly, Serum miR-21 degrees were important addition in hormone-refractory prostate malignant neoplastic disease patients, which could be used as a possible forecaster for the docetaxel-based chemotherapy [ 24 ] . Furthermore, miR-21 was significantly over-expressed in tumour tissues compared with those non-tumor tissues, bespeaking that miR-21 in serum was a possible biomarker for chest malignant neoplastic disease diagnosing [ 25 ] . Additionally, in non-small-cell lung malignant neoplastic disease, miR-21 besides showed a high sensitiveness of 91.67 % in diagnosing of the malignant neoplastic disease, and Shenet Al.concluded that miR-21 look in plasma could be a possible blood-based biomarker for NSCLC [ 26 ] . Above all, old surveies identified miR-21 as a sensitive and specific blood-based biomarker in assorted malignant neoplastic diseases. However, meta-analyses got different decisions approximately miR-21 as a blood-based biomarker. Yang et Al. performed a meta-analysis ( including 29 articles up to March 2012 ) to measure the sensitiveness and specificity of miR-21 look in diagnosing of NSCLC. They indicated that miR-21 may non be suited as a diagnostic biomarker for NSCLC because the degree of its diagnostic truth ( sensitiveness of 75.7 % and specificity of 79.3 % ) was non high plenty for clinical pattern [ 27 ] . The same decision was reached by present survey. In this meta-analysis, we concluded that look of miR-21 in blood cells had small association with the presence of NSCLC. The ground for this consequence may be that miR-21 look was different in blood cells and plasma.
Some restrictions were existed in present meta-analysis. ( 1 ) Significant heterogeneousness was observed in this survey. On the one manus, blood cell samples included in four eligible datasets were either peripheral blood ( GSE24709 and GSE17681 ) or whole blood ( GSE27486 and GSE40738 ) , which would ensue important heterogeneousness. On the other manus, blood cell samples showing miR-21 belonged to different states ( German and American severally ) , in which people have different life wonts. ( 2 ) The low figure of cistron look datasets limited farther analyses such as stratification analysis and publication prejudice analysis, which leaded to incomplete appraisal. ( 3 ) As the information included in selected datasets were uncomplete, factors such as gender and age were excluded in this meta-analysis. ( 4 ) Because of the little sum of incorporated informations, the consequences obtained through the meta-analysis were unstable. More datasets and samples should be chosen to measure the suitableness and effectivity of miR-21 as a blood-based biomarker in early diagnosing of NSCLC.
The present survey was the first meta-analysis to measure the effectivity of miR-21 as a blood-based biomarker in early diagnosing of NSCLC by utilizing the microarray datasets. It was concluded that sensing of miR-21 look was non suited for the early diagnosing of NSCLC and miR-21 was uneffective as the blood-based biomarker in NSCLC.