Biomarker Intelligent Mining Of Pancreatic Ductal Adenocarcinoma Based On Omics Data

Cancer is a general term for a large group of diseases caused by the uncontrolled growth and division of certain cells in the body and their spread to other parts of the body.As one of the highly aggressive and lethal solid malignant tumors,pancreatic ductal adenocarcinoma has no obvious symptoms in its early stages and the corresponding symptoms do not appear until the patient reaches the advanced stage of cancer,making it difficult to be diagnosed in the early stages of cancer.In addition,its highly heterogeneous intra-tumor characteristics make the effectiveness of clinical treatment greatly compromised.Although there has been an increasing number of studies related to it in recent years,the understanding of the etiology and diagnosis and treatment of pancreatic ductal adenocarcinoma is still insufficient.Therefore,it is necessary to study the subtyping of pancreatic ductal adenocarcinoma to dissect its intra-tumor heterogeneity,and to use computer technology to mine the important information hidden in the sample data to identify potential biomarkers that can be used as early diagnosis and clinical treatment of pancreatic ductal adenocarcinoma.In this paper,we first developed a cancer subtyping algorithm based on gene co-expression network alignment and machine learning algorithms and identified biomarkers associated with pancreatic ductal adenocarcinoma subtypes;Furthermore,we successfully constructed a weighted gene co-expression network for pancreatic ductal adenocarcinoma based on complex network theory and identified immune-associated biomarkers;Finally,based on the genes of immunogenic cell death(ICD)-related damage-associated molecular patterns(DAMPs),we successfully constructed a prognostic risk model for pancreatic ductal adenocarcinoma using machine learning algorithm and mined biomarkers associated with immune prognosis of pancreatic ductal adenocarcinoma.The main research contents and related results are as follows.In the first part,a cancer subtype typing algorithm based on gene co-expression network comparison was developed and named GCNA-cluster.First,the algorithm constructs a weighted gene co-expression network for each patient.Second,the gene co-expression network of every two patients is compared,which takes into account the synergistic effects between genes.A scoring function is also defined to measure the results of the network comparison,and the scores of the network comparison can indicate the similarity between patients.Then,patients are clustered according to their similarity.The accuracy of the algorithm was also validated on a pancreatic ductal adenocarcinoma dataset from the GEO database with real labels,and the experimental results showed that the GCNA-cluster algorithm had better results than the classical cancer typing algorithm.Next,the GCNA-cluster algorithm was applied to the pancreatic ductal adenocarcinoma dataset from the TCGA database to identify two cancer subtypes and evaluate them using silhouette coefficients.In addition,the key genes of each subtype were screened by overlaying the maximum concatenated subnetworks generated by network matching,several biomarkers related to pancreatic ductal adenocarcinoma subtypes were identified that could regulate cell growth,cell cycle or apoptosis and provide new therapeutic strategies for clinical practice.In the second part,lnc RNA-m RNA co-expression network was constructed to screen and identify immune-related biomarkers of pancreatic ductal adenocarcinoma.First,the significantly differentially expressed lnc RNAs and m RNAs were screened based on differential analysis,and the WGCNA algorithm was used to construct the lnc RNA-m RNA co-expression network for the differentially expressed genes.Then,the gene expression data of pancreatic ductal adenocarcinoma samples were immunoassayed using the inverse convolution algorithm CIBERSORT,and the results of immuno-infiltration analysis were used as external features of the lnc RNA-m RNA co-expression network.Finally,the gene co-expression modules that were highly correlated with the initial B cells were screened based on the similarity of gene expression,and topological analysis and pathway enrichment analysis were performed.It was found that m RNAs ranked higher in the network in terms of degree,mediator centrality and near-centrality,which were the hub genes in the co-expression network,and these hub genes were all enriched in B cell-related immune pathways,with which co-expressed lnc RNAs were presumably associated with pancreatic ductal adenocarcinoma immune mechanisms,and these genes could be used as potential immune biomarkers for pancreatic ductal adenocarcinoma and provide a reference for immunotherapy of pancreatic ductal adenocarcinoma.These genes can be used as potential immune biomarkers for immunotherapy of pancreatic ductal adenocarcinoma.In the third part,the analysis based on immune genes was used to construct a prognostic risk model for pancreatic ductal adenocarcinoma and identify prognostic immune-related biomarkers.First,based on 32 DAMPs-related genes,a consensus clustering algorithm was used to classify pancreatic ductal adenocarcinoma into two DAMPs-related subtypes.Mutation analysis of these two subtypes revealed that the mutation frequencies of KRAS,TP53,SMAD4 and CDKN2A were lower in subtype C1 than in subtype C2.Then,one-way Cox regression analysis and LASSO Cox regression analysis were performed on genes differentially expressed between subtypes to construct an 18-gene prognostic risk model for pancreatic ductal adenocarcinoma.Based on this model,pancreatic ductal adenocarcinoma samples from TCGA database and ICGC database were divided into high and low risk groups,respectively,and survival analysis and immunotherapy response detection were performed for high and low risk groups.The 18 prognostic immune-related biomarkers obtained from the experimental screening are useful for health monitoring,survival prognosis and immunotherapy of pancreatic ductal adenocarcinoma patients.In this paper,we use machine learning algorithms to comprehensively analyze biomarkers related to pancreatic ductal adenocarcinoma subtypes,immune correlation,and prognostic immune correlation from multiple perspectives,and these diagnostic,prognostic,and therapeutic biomarkers will help clinical workers to develop more personalized treatment strategies for patients and the development of targeted cancer drugs.