Construction And Analysis Of Specific Gene Regulatory Networks For ESCA And STAD

Complex diseases are a major challenge in human medicine and biology in the 21 st century.These diseases are dominated by both genetic and environmental factors.There are many genetic materials involved in the occurrence of complex diseases,such as genes,proteins and so on.The transcription process from DNA to RNA is an important step in the expression of genetic material.Protein phosphorylation,as an important post-translational modification method,has a significant impact on the function of proteins.The study of protein phosphorylation and transcriptional regulation is helpful to further understand the mechanism and essence of life activities.Stomach adenocarcinoma and esophageal carcinoma are two complex diseases with very hign mortality.Most of the studies on their molecular mechanisms and cancer biomarkers are based on proteomic data.Combining with genomic data and proteomic data,the study of cancer biomarkers for stomach adenocarcinoma and esophageal carcinoma from the perspective of network gets less attention.Therefore,in this thesis,we intend to integrate expression data and methylation data to construct two cancer-specific network for stomach adenocarcinoma and esophageal carcinoma,and identify cancer-specific modules using weighted network analysis,and further predict potential cancer markers within the modules.Firstly,considering the important role of methylation in carcinogenesis,abnormal methylation molecules were extracted from two cancers.As a candidate set of potential cancer biomarkers,it is also a node set of disease-specific networks.Besides,we integrate the genome-wide regulatory relationship data,and the relationship data between kinases and substrates in phosphorylation modification.In these relational data,methylation abnormal molecule is chosen as the endpoint and the correlation coefficient between the two endpoints is large,as the edge set of disease-specific network.Then,in this thesis,weighted co-expression analysis is carried out on two specific networks.Hierarchical clustering algorithm is applied to the two networks,and the network is dynamically cut into modules of different sizes.The stomach adenocarcinoma network is divided into 14 modules,and the esophageal carcinoma network is divided into 10 modules.Based on the idea of dimensionality reduction,a new method for predicting specific modules and potential cancer biomarkers is proposed.Taking stomach adenocarcinoma as an example,the correlation coefficients between the first principal component of each module and all stomach adenocarcinoma markers were calculated,and the modules with large difference in correlation coefficients were selected as specific modules to predict potential cancer biomarkers.Among the two cancers,the top three modules were selected for further analysis.Finally,in weighted special networks,distance-based and correlation-based similarity evaluation methods can be used to measure the relationship between molecules.Therefore,three distance-based methods and two correlation-based methods are used to rank the molecules in each module and the first principal component of the module.The highest prediction accuracy is analyzed and compared in this problem.The experimental results show that the correlation-based method is better than the distance-based method in the prediction of disease markers for gastric and esophageal cancer,and the Pearson correlation coefficient is more prominent in the correlation-based method.Among the six specific modules,an average of 32% of the molecules can be verified.In this thesis,we constructed and analyzed the stomach adenocarcinoma specific network and esophageal carcinoma specific network.We deeply analyzed the cancer specific modules presented in the two networks,used various methods to predict potential cancer biomarkers,and compared the results of different methods.Gene enrichment analysis of the module was carried out,and the two cancers were analyzed from the perspective of biochemical pathway.However,the histological data used in this thesis are limited.In the future,more high-throughput data such as somatic cell mutation data can be integrated to construct a more perfect specific network,which will contribute to life science and target drugs as well as precise medical treatment.