Cooperativity Of Transcription Factor Binding And Its Regulation Role On Gene Expression

Transcription factors play a key role in the process of regulating gene expression.Transcription factortend to localize in particular DNA sequences to achieve complex and accurate regulation of target genes.Actually,different transcription factors always regulate gene expression by interaction patterns.In order to survey the interaction of transcription factors related with gene expression level,two gene sets with high expression level and low expression level have been established utilizing two RNA-seq data of GM12878 cell line.Based on the ChIP-seq data of 83 transcription factors downloaded from ENCODE,the TF association strengths(TFAS)of 83 transcription factors are calculated.And their interaction networks are obtained in both gene sets and visualized using Cytoscape.By adjusting the thresholds we get an important transcription factor BCL11 A and a specific interaction mode NFYA-NFYB-SP1 in high expression gene interaction network.We found that some transcription factors tend to form submodule to regulate gene expression.At last,we obtained three submodules(D,E,F)only presented in high expression gene set and three submodules(A,B,C)presented in both gene sets.Then,the signal of all TF in the above submodules were used to predict gene expression level utilizing CNNs(Convolution Neural Networks).The results indicated that the non-specific submodule A(ATF2,BATF,IRF4,BCL11 A,NFIC,FOXM1)has the best performance(AUC average=89.33%),and the high-expression specific submodule F(TCF12,TCF3,PML,TAF1,POU2F2)has the second performance(AUC average=87.56%).Finally,we integrated the signals of each submodules and the fusion feature of each submodule was used to predict gene expression level.We found that the AUC value was increased for three high-expression specific submodules but the AUC value was decreased forthe other three non-specific submodules.The results further indicated that there are imitate relationship between these transcription factors in three submodules with high expression specificity.