Prediction Of MiRNA Regulatory Modules With Human Non-small Cell Lung Cancer By Rule Induction

IntroductionMicroRNAs (miRNAs) as a class of non-coding RNA regulate gene expression at post-transcription level. More and more studies reported that miRNAs were associated with multiple cancers and played the important roles in carcinogenesis, progression and metastasis. However their regulation progress, involved in pathways and the relationship between them and their target genes are still less understood. In order to learn the regulatory mechanism of miRNAs and the interactions between miRNAs and target genes in the development of cancer, we try to find their functional modules in cellular systems.Materials and methodsOur study downloaded the miRNA expression data E-TABM-22 from the Arrayexpress of EBI (www.ebi.ac.uk) used-"lung cancer" and "microRNA". The data set includes 245 individual samples of lung cancer and lung cancer cell line, which contain 104 samples of non-small cell lung cancer (65 adenocarcinoma and 39 squamous cell carcinoma) expression data. Gene expression data sets which used in this study are GSE2514, GSE3141, GSE4573, GSE5123, GSE5843 and GSE6253. Data set of miRNA target gene combined with from the mirna.org (www.mirna.org) download. And then we selected the miRNAs and genes which were overlapping of miRNAs or genes from three kind of dataset to calculate the corresponding correlation matrixes. We calculated and established one miRNA's correlation matrix and eight gene's correlation matrixes by correlation analysis with miRNAs or genes. We induced MRM by analyzing miRNA, target gene expression matrix and the miRNA-target gene regulation with CN2-SD method. Rule induction form is IF [Cond] THEN [Class Distribution].8 gene matrixes were respectively calculated rule sets according to 0.1,0.2,0.3,0.4 and 0.5 correlation coefficient level. And then we removed redundant rules, merged rule sets, finally deleted the rule which was induced by only one gene data set.ResultsAfter rule induction, we evaluated the induced MRM by using the confidence and coverage. In order to ensure the accuracy of induced rules, we selected MRMs as an effective rules at their confidence≥0.6 and coverage≥3. The higher is the correlation coefficient, the fewer number of target genes the MRM included. After merged rule set, we finally got MRMs with non-small cell lung cancer. At PCC=0.2, confidence≥0.75 and coverage≥3, we got 166 MRMs. In this study, MRMs were divided into three categories in accordance with the histological type of gene expression data. I type were induced from genes matrixes which the histological type of sample is adenocarcinoma,Ⅱtype were the genes matrixes of squamous cell carcinoma; III category were induced from both adenocarcinoma and squamous cell carcinoma.We also validated MRMs (PCC=0.2) by GOstat software. In this study,119 GO annotation entries were selected, in which 66 related to biological processes,17 related to cellular components,36 related to cthe molecular function.DiscussionThe miRNAs have multiple functions in lung development, and abnormal expression of miRNAs could lead to lung tumorigenesis. The miRNAs act two rules, oncogenes (e.g. miR-17-92 cluster) and tumor suppressor genes (e.g. let-7 family). Disrupted let-7 function is a feature of human lung cancer. There were 31 MRMs associated with let-7 family in 166 MRMs (PCC=0.2). The genes in MRMs have the following functions:biological process positive regulation of protein metabolism, effects of enzyme and ATP hydrolysis activity (GOstat). The genes in module 14, module 19 and module 34 were associated with non-small cell lung cancer. miR-17-92 cluster overexpression play an important role in lung cancer formation and progression. The main features of genes in module 111 are:negative regulation of nucleotide metabolism (ATR and ATXN1), DNA damage repair (ATR and ATRX). The genes in module 60 and module 63 were associated with tumor metastasis and cell differentiation.In our study, an improved method was used to finding the MRM with non-small cell lung cancer. We changed the selected data source, downloaded data from the standard database. And we increased the filter rules, that is, to delete the rule which was induced by only one gene data set. After improving the method, this study got smaller MRM data set, but increased the biological significance of MRM. Moreover MRMs with non-small cell lung cancer were consistent with biological processes which were reported in the recent literature. Therefore, our method can be used to find meaningful miRNA and target gene set.The MRMs can support the hypothesis for further clarify the non-small cell lung cancer mechanism associated with the miRNA. The miRNAs and genes in MRM can also be as the biomarkers for the diagnose of non-small cell lung cancer or the target for the gene therapy of non-small cell lung cancer. The special MRM can be as subtype classification of the non-small cell lung cancer.ConclusionThis study found the MRM related to non-small cell lung cancer by means of rule induction.After verifying the MRM by literature and GOstat, the induced MRM in this study has biological significance, so our method can be used for finding MRM related to tumor.