Investigation On Systems Biology Of Molecular Mechanism Involving In Pathogenesis And Immunity Of Avian Influenza Virus

Influenza virus is a negative strand RNA virus, it may cause severe influenza pandemicin the whole world. Integrative approach combining gene expression data and static proteininteraction networks has been rapidly developed, which is more suitable for analysis ofspecific biomedical problems. Methods treating biological pathway as biomarkers have beenproposed, which have higher robustness and accuracy compared to methods based ondifferential expression genes for selecting biomarkers. Pathway interaction analysis canrecognize a disease process from a new perspective.In this study, GSE2713, a dataset of human infected with H1N1, was used as goldstandard dataset. We compared the performance of five methods based on the selected goldstandard dataset, which are Significance Analysis of Microarrays (SAMVB), Gene SetEnrichment Analysis (GSEA), Principal Component Analysis (PCA), Bioconductor, andIdentifying Different Expression Genes From Pathway Interaction Networks (PIN),respectively. Results showed that the PIN method has the highest accuracy,80.6%,meanwhile has a higher sensitivity,81.3%. Furthermore, the five-fold cross-validation methodwas used to test the reliability of the five methods. And results showed that the PIN methodhas the maximum area under curve,0.75, which means that this method has the highestreliability.We chose four datasets of chicken infected with avian influenza virus from GEOdatabase, which were GSE32378, chickens infected with H5N1; GSE31476, chickensinfected with H1N1; GSE31505, chickens infected with H9N2; GSE31506, chickens infectedwith H5N2. First, we applied the selected PIN method to filter dysregulated pathway anddifferent expression genes. Then, we studied different genes from such prospective:chromosomal localization, GO annotation enrichment analysis, mapping to protein-proteininteraction network. Results showed that there is small difference in the distribution ofchromosomal location, and there is no significant difference of their GO annotationenrichment analysis. However, protein-protein interaction network topology analysis showedthat the four datasets possess different hub proteins, which means there are different important genes in these datasets. In addition, we filtered different important genes by analysing theinteraction network of pathways in different flu virus infection process. Our results showedthat, CBL and GOT1play an important role in chickens infected with H1N1or H5N2;ATP2A2is filtered as an important gene in chickens infected with H9N2or H5N2. Ribosome,oxidative phosphorylation, purine metabolism, wnt signaling pathway, TGF-beta signalingpathway and Jak-STAT signaling pathways are always important in the four infectionprocesses.Besides, we chose three datasets of different species infected with H1N1avian influenzafrom GEO database: GSE31524, human infected with H1N1; GSE35738, pigs infected withH1N1; GSE31476, chickens infected with H1N1. Based on the same analysis process, we cansee that, there are significant differences existed in the biological process, molecular functionand cellular component GO annotation enrichment analysis; from the topology analysis ofprotein-protein interaction network, the different hub proteins obtained from the three datasets,which mapped to different import genes; and different important genes were screened byanalysis of interaction network of pathways as well. All these results indicate MAPKsignaling pathway and Jak-STAT signaling pathway play an important role during H1N1infection.