Research On The Inference Method For Large-scale Gene Regulatory Network Model

Gene regulatory network is a kind of biochemical network composed of the interaction among genes.Gene regulatory network enables biologists to understand the regulatory relationship among genes and the highly complex biological phenomena from a systematic perspective,which plays an important role in the research of complex diseases such as tumors.A regulatory network composed by a few genes can usually be obtained from classical experimental methods.But biological experiments usually take a lot of time and resources,so it is becoming an important method by using artificial intelligence,signal processing and other new technologies to infer gene regulatory network.With the development of next generation sequencing technology,researchers have obtained a huge number of gene expression data,which provide the foundation for the construction of gene regulatory network.With the development of information science,many different mathematical models for gene regulatory network have been proposed.Differential equation model is a dynamic mathematical model,which is suitable for reflecting the process of gene expression concentration level changing over time.It can extract the specific interaction among genes from time series data.In this paper,we improve the inference algorithm of gene regulatory network based on differential equation model,and propose an improved inference algorithm,which is more robust than the existing algorithm.Furthermore,it is combined with gene pre-selection method on this basis,and the ability of the algorithm to infer large-scale gene regulation relationship is extended.Finally,the effectiveness of proposed method is verified by bioinformatics analysis.The main contents of this paper are as follows:(1)Improvement of algorithm.The model inferred by genetic programming combined with filtering algorithm has higher complexity and unstable results when the number of genes is larger.In this paper,genetic programming is improved by controlling the order of model,reducing the complexity of the model,and enhancing the stability of the results.And the performance of the improved joint algorithm is analyzed by experiment.(2)Pre selection of regulatory genes.The selection algorithm is applied to screen the regulatory genes,reduces the number of candidate genes in the process of model inference,it expands the application of the inference algorithm for large-scale gene regulatory network,and improves the accuracy of the inference models.(3)The application of the inference algorithm for large-scale gene regulatory network in biomedicine.The yeast dataset is employed to test the joint algorithm,and the effectiveness of proposed algorithm is verified by comparing the results of biological experiments.At last,we infer the large-scale gene regulatory network model by using human cervical cancer dataset,and analyze the relationships among genes and their impacts on complex diseases.