Application Of Data Mining In The Investigation Of Transcription Of NDRG2and NDRG2-mediated Signaling Pathway

The recent high throughput techniques used in biomedical research haveresulted in a flood of biological data, ranging from genomic and proteinsequences, DNA microarrays, to biomedical images. To exploit these data fordiscovering knowledge that can be translated into medical research application,data mining has been designed and developed to handle data analysis problems. Itplays an increasingly crucial role in revolutionizing biomedical research andbecome integrated into the pipeline of biomedical discovery process. In thisthesis, we employed data mining skills to extract meaningful information fromexpression profile of HepG2cells, predict the transcription factors regulating thetranscription of NDRG2and the interactants of NDRG2and Dusp6.Since NDRG2expression is downregualed in cancer cells, to provide aglobal view of biological effect of NDRG2on tumor cells, we first conducted amicroarray study to detect the expression profile of HepG2cells and performedenrichment analysis on microarray data. Gene Ontology (GO) biological processanalysis revealed that genes related to G protein signaling pathway wereupregulated. Five of them were selected and verified by real-time PCR. Genesrelated to M phase of cell cycle were downregulated. This was in agreement with cell cycle analysis. Signaling pathway analysis demonstrated apparent augmentedhematopoietic cell lineage pathway and cell adhesion, but reducedGlycosylphosphatidylinositol (GPI)-anchor biosynthesis, protein degradation andSNARE interactions. Furthermore, through motif analysis and experimentalvalidation, we found that the p38phosphorylation can be increased by NDRG2.Through enrichment analysis, we successfully extracted meaningful informationfrom microarrays, and provided the molecular basis for understanding the role ofNDRG2in tumor cellsIn order to understand the expression pattern of NDRG2under differentconditions, we predicted the transcription factors regulating NDRG2expressionby combination of ARACNE algorithm and motif scan. Through motif scan,129putative transcription factors binding site were identified in NDRG2promoterfragment. To improve prediction precision, ARACNE algorithm was introducedto estimate the correlation between NDRG2and the transcription factorcandidates. Finally, we achieved a list of53factors potentially regulatingNDRG2expression. Among these factors, KLF4was selected for furtherexperimental validation for its role in promoting differentiation of colon cancercells. Functional annotation of transcription factors revealed that the enrichedcategories of these factors were linked to cell differentiation, organ developmentand cellular localization, which were consistent with previous studies.Finally, To find the interacting partner of NDRG2, we looked for thehomologs of NDRG2in Arabidopsis thaliana, and identified NDL1, NDL2andNDL3as homologs of NDRG2. NDL proteins have been reported to interact withAGB1and RGS1in Arabidopsis, which provide us the clue that NDRG2mightinteract with homologs of AGB1or RGS1in human. AGB1is a β subunit of Gprotein heterotrimer. Five homologs (GNB1-GNB5) of AGB1were identified in human. RGS1is a member of GTPase activating proteins, which directlyinteracts with α subunit of G protein heterotrimer to catalyze GTP hydrolysis.Although only one RGS protein (RGS1) was found in Arabidopsis, nearly twentyRGS proteins were identified in human, though they don’t have remarkablesequence similarity with RGS1from Arabidopsis. Among these proteins, humanRGS5has the most sequence similarity with RGS1, and was selected for furtherexperimental validation. By Co-IP and His-pulldown assay, we identified RGS5as an interactant of NDRG2. To establish a method to predict the interactants of aparticular protein, we made use of gene expression pattern and protein sequencefeatures to predict the new interactants of Dusp6. MINDy algorithm wasintroduced to find the modulator of Dusp6, and Pred_PPI was used to improvethe precision of prediction. By this method, we predict Mapk8as an interactant ofDusp6, which was confirmed by immunoprecipitation.Taken together, through combination of data mining and experimentalvalidation, we successfully established a method to investigate the transcriptionof a particular gene and its signaling pathway. As these methods are time-savingand cost-effective, we anticipate that they will be applied to the investigation ofother genes.