Comparative Analysis Of The Molecular Mechanism For ESCA And STAD Based On Gene Regulatory Network

Among all kinds of diseases,cancers have a huge impact on human health due to their high mortality and high recurrence rate.Esophageal cancer(ESCA)and stomach cancer(STAD)which have poor prognosis are two common types of malignant tumors in digestive tract.It plays a key role in understanding the molecular mechanism for ESCA and STAD.Gene expression regulation endows cells with structural and functional control,which is the basis of cell differentiation,morphogenesis and the versatile and adaptive functions of any organism.There are a variety of regulatory modes in organisms that involved by multiple biomolecules.Micro RNAs(mi RNAs)and transcription factors(TFs)are two important regulatory factors among these biomolecules.In particular,a feed-forward loop(FFL),composed of a mi RNA,a transcription factor and their common target gene,is an important regulatory mode and has been proved to be relevant to the development and progression of complex diseases,including cancers.In this thesis,we proposed a biomarker prediction framework based on the gene regulatory network to explore the molecular mechanism for diseases.Firstly we performed data preprocessing,including expression data,regulatory data and so on.And then we used the differentially expressed molecules,regulatory data and expression data to calculate the similarity between the biomolecules which had the regulatory relationship,thus we constructed a disease-specific gene regulatory network.Three typical types of three-node FFLs were identified from the disease-specific gene regulatory network to obtain the disease-specific feed-forward loop.A disease-specific gene co-expression network was then constructed based on biomolecules in the disease-specific feed-forward loop.The random walk with restart prediction method was performed in the disease-specific gene coexpression network to obtain a scoring list of candidate biomolecules,and the higher the score was,the more relevant the biomolecule was to the specific disease.Based on the proposed framework,we analyzed ESCA and STAD.We identified 148 threenode FFLs which were made up of an ESCA-specific feed-forward loop and 242 three-node FFLs which were made up of a STAD-specific feed-forward loop.Enrichment analysis of genes and transcription factors in the ESCA-specific feed-forward loop and STAD-specific feed-forward loop showed that these biomolecules were involved in the development of diseases.At the same time,the same part of these two disease-specific feed-forward loops were analyzed,and the same part of the two disease-specific feed-forward loop’s enrichment results were also analyzed.These two steps showed that ESCA was associated with STAD.The observation method was used to obtain the threshold of the construction of the diseasespecific gene co-expression networks for ESCA and STAD.And we used the leave-one-out cross validation method to evaluate the prediction method,and when the restart probability was gradually increased from 0.1 to 0.9,the AUC average was also increased based on these two types of diseases-related data.Through the prediction method,two scoring lists of candidate biomolecules were obtained.And we analyzed the candidate biomolecules whose scores were ranked top 20,and discovered there were 12 and 13 biomolecules related to esophageal cancer and stomach cancer respectively.We also analyzed the biomolecules which were ranked closer and higher.These biomolecules had same expression trends,and the gene KIT was related with both of these two types of cancers.Multiple analyses showed that the biomarker prediction framework proposed in this thesis effectively predicted the disease-related biomolecules for ESCA and STAD and discovered that there was a certain correlation between these two types of cancers,which helped develop diagnostic and therapeutic strategies for esophageal cancer and stomach cancer.