| Gene regulatory network(GRN)refers to the network formed by the interaction between genes and their expression products during the process of gene transcription and expression in cells during biological growth and development.Although people can only identify some of these relationships in real experiments,they cannot reconstruct a complete GRN.Therefore,based on experimental measurements of gene expression and partial regulatory relationships in organisms,using computers to explore undiscovered correlations between genes,integrating and reconstructing GRNs,can explore regulatory relationships between genes at the gene level,which is of great significance for revealing gene function,studying cell transformation,and exploring the relationship between biological genotypes and phenotypes.The high noise and dimension of transcriptome data and the high sparsity of gene regulation relationship make it extremely difficult to reconstruct GRN.To address the above issues,this article proposes a GRN reconstruction method based on pseudo twin networks by constructing training data reasonably and analyzing temporal and spatial features of gene expression profiles using pseudo twin networks.The robustness and effectiveness of the model are also studied.This article proposes a PSGRN model based on the maize seed development gene expression profile and maize leaf GRN dataset,which can be used to reconstruct large-scale GRN using time series expression datasets.This framework captures the temporal and spatial characteristics of transcription factors(TFs)and target gene(Target)expression sequences through the Pseudo Siamese Network,to analyze the correlation between each,and finally evaluates the interaction between the two through the sigmoid function to reconstruct the entire GRN.The experimental results show that under the same data volume and gene expression feature data conditions,the PSGRN model performs better than the current state-of-the-art methods,while also possessing robustness and generalization.The study also validated the effectiveness of each module in PSGRN and demonstrated the rationality of the model.This article uses the PSGRN model to predict gene interaction relationships on the maize gene expression dataset,and further analyzes the top 15 TFs in maize GRN and their regulatory effects on gene expression.Research has found that these TFs are highly expressed in the process of double fertilization and co fertilization of maize seeds,and have been associated with other genes.In addition,this article also conducted interaction network analysis using the STRING database,confirming that the PSGRN model can effectively identify regulatory relationships existing in the network and infer new potential regulatory relationships.These results provide a basis for supplementing biological data and determining key regulatory factors for maize differentiation,and provide guidance for further research on genotype phenotype associations in maize. |
