Comprehensive Network Identification Of Key Genes In Pancreatic Cancer

Nowadays,cancer threatens people’s health all the time,and most cancers have a poor prognosis,and the mortality rate of pancreatic cancer is even close to the incidence rate.Early diagnosis and precise treatment are critical to improve the survival rate of pancreatic cancer patients.But so far,the molecular mechanism of pancreatic cancer occurrence and the perfect and reliable methods for the diagnosis and treatment of pancreatic cancer has not been accurately grasped.Humans have a large number of genes,and it is unrealistic to quickly and comprehensively study the impact of each gene on the occurrence of pancreatic cancer.Therefore,it is urgent to screen out the genes that play a key role in the development of pancreatic cancer and conduct in-depth research on them.With the advent of the information age,machine learning and deep learning has also been widely used in the field of bioinformatics.In this paper,based on a variety of machine learning and deep learning algorithms,a comprehensive network model for identifying the key genes of pancreatic cancer has been obtained.It is hoped that the selected genes can become the "crucial keys" to accurately study the molecular mechanism of pancreatic cancer,and can be used as valuable clinical markers and therapeutic targets for pancreatic cancer.This paper proposes a four-step comprehensive network method named Bio-SVMRFE-DNN（Biological SVM-RFE Deep Neural Network）for the screening of key genes in pancreatic cancer.Firstly,the 7 gene expression profile files from the GEO database are filtered,standardized and corrected for batch effects,and then performs gene differential expression analysis on the processed data,and 646 stable different genes for pancreatic cancer are identified using the Robust Rank Aggregation algorithm.Next,37 feature genes are preliminarily obtained by using the SVM-RFE algorithm based on the recursive feature deletion criterion designed in this paper.Then,through Weighted Gene Co-Expression Network Analysis,15 core genes that play an important role in the protein level are obtained from the list of characteristic genes.Finally,through the deep neural network containing three hidden layers and the backpropagation feature selection method IBFS proposed in this paper,we identified 7 key genes of pancreatic cancer: SDR16C5,AHNAK2,LAMC2,CTSE,TSPAN1,S100 P and TMPRSS4.This paper compares the Bio-SVM-RFE-DNN the three feature selection methods:Random Forest,SVM-RFE and LASSO regression.The results show that the method proposed in this paper has the highest classification accuracy and can provide medical personnel with a smaller and more accurate research range.After consulting various literature,it was found that most genes（such as SDR16C5,S100P）have been confirmed to play an important role in the development of pancreatic cancer.Early detection of the content of these genes is helpful for the early diagnosis of pancreatic cancer.In addition,this paper also conducts functional analysis,different analysis,survival analysis,and Cox analysis of the screened genes.The results show that these genes are closely related to the survival of pancreatic cancer patients,and their abnormal expression will affect the development of cancer.A large number of experimental results prove that the comprehensive network for identifying key genes of pancreatic cancer designed in this paper is ideal,and can provide valuable reference information for the study of the development mechanism and precise treatment of pancreatic cancer.