The Research On Algorithm Of Identifying Master Regulators Based On The Co-regulatory Network

In bioinformatics,the identification of master regulators is a very important topic.Transcription factors and miRNAs are two important regulators in eukaryotic cells that regulate gene expression at the transcriptional and translational levels,respectively.At the same time,there are also a wide range of interactions and co-regulations between transcription factors and miRNAs.These relationships constitute a co-regulatory network.Therefore,studying the mechanism of master regulators in the co-regulatory network plays a crucial role in understanding the biological processes and mechanisms of disease.To this end,we propose two master regulators recognition algorithms in this thesis based on the related information such as co-regulation network and expression profile data.First of all,the traditional master regulators recognition algorithms mainly use the expression profile data,and the expression profile data has certain noise.In addition,some methods only consider a single regulatory network,ignoring the complex regulatory relationships in the co-regulation network.Therefore,in this thesis we propose a master regulator recognition algorithm MRCE based on co-regulation relationship and expression profile data.According to the expression profile data,the algorithm first integrates the co-regulation relationship and the information of key gene to calculate the regulatory scores of the regulators.Subsequently,the Pearson correlation coefficient is used to calculate the interaction between the regulators.Then,based on the Kendall coefficient,the consistency between the regulatory scores of the regulatorrs and the correlation coefficients is evaluated.Finally,in order to reduce the influence of noise in the expression profile,MRCE is combined with bootstrap for repeated sampling,and the regulators with p_value less than a certain threshold are selected as the candidate master regulators.The experimental results show that MRCE can effectively identify master regulators with significant biological significance in breast cancer and thyroid cancer.Secondly,due to the sparse of co-regulatory relationships,the traditional algorithms based on regulation relationship have certain false positives.In addition,the serial algorithms have a large search space and low time efficiency.In view of the above problems,we propose a master regulator recognition algorithm MRDGC based on the ability of differential expression of gene and regulation ability of regulators in this thesis.The algorithm firstly pre-processes the genes and selects the genes with obvious differential expression.Then,in order to avoid the influence of the sparse of co-regulatory relationships,MRDGC calculates the regulatory ability score of the regulators as intensity of regulatory factors on gene regulation.Finally,integrating the ability of gene differential expression and regulators to control the dis tribution of regulators we select the meaningful regulators in the test as candidate master regulators using parallel Wilcoxon rank sum test,and then identify top k master regulators from candidate master regulators.Through experimental comparison,functional enrichment analysis,parallel result analysis,robust analysis and case study,all the results show that MRDGC has high stability and efficiency.