Analysis About Transcription Targets Of Embryonic Stem Cells

Transcription factors(TFs)OCT4,SOX2,and NANOG play a vital role in the embryonic stem cells'(ESCs)self-renewal and differentiation,known as ESCs' pluripotency maintaining ‘core' transcription factors.However,the regulation characteristics and rules of the ‘core' transcription factors on target genes are still unclear.Besides,the lack of target genes also restrict the exploration of ESCs' self-renewal and differentiation mechanisms.It is still unclear whether there are interactions and coordinated regulations between these transcription factors.ESCs-specific microRNAs(miRNAs)play an important role in the maintenance and differentiation of ESCs' pluripotency.Therefore,exploring regulation factors' transcriptional and post-transcriptional regulation will help us to explore the molecular mechanism of pluripotency maintenance of ESCs.There are still many difficulties in the mining of ‘core' TFs' targets due to social ethical issues,which leads to a relatively few number of human target genes compared with mouse and affects further molecular experimental studies.Therefore,in this study we explored the regulation features of the ‘core' TFs' targets in ESCs by using system biology methods and predicted the ‘core' TFs' targets by using the genomic sequence features and epigenetic regulation characteristics based on the machine learning methods.In addition,we also extracted and screened human cell-specific molecular networks and calculated the cellto-cell similarity,which provides theoretical support for experimental verification of downstream molecules.The main findings are as follows:(1)PPIN(BioGRID,HPRD)and miRNAs regulation networks(miRecords,TarBase)are integrated to analyze topological properties of ‘core' TFs' targets.It was found that the ‘core' TFs and miRNAs co-regulated genes are significantly different from the ‘core' TFs solely regulated genes and shows more significant difference comparing with other genes in both human and mouse PPINs.This result indicates that ‘core' TFs and miRNAs synergistically regulate and enhance target genes' function in order to achieve their goal of maintain ESCs' pluripotency.(2)We built a negative set needless model named Label method algorithm(LMA)and used it to predict ‘core' TFs target genes in human.Histone modification features and genome-wide transcriptional factors regulation features were used in the prediction.We predicted 4796,3166,and 4384 target genes for OCT4,SOX2,and NANOG respectively and offered reliability scores each predicted targets.Compared with a previous mapping-convergence(M-C)algorithm,LMA model has a higher stability and a positive set accuracy,which lays a solid foundation for further researches.Predicted targets indicate strong modularity and gene function similarity through analyzing in gene ontology database and PPINs,which confirms the accuracy of LMA model predicted results.(3)We integrated all cell types and cell regulatory networks in human and built a cell similarity calculation and cell type prediction software named CellSim.CellSim can calculate the similarity of all human cell types and provide shared detail regulation networks of some cells with reliability scores,which help researchers to select specific TFs for cell direct reprogramming.Besides,CellSim can calculate matched cell types based on specific gene sets in unlabeled cells.For the convenience of users,the calculation results of this software are available for downloading.