Construction Of Prognostic Prediction Model And Target Genes Screening Of Muscle Invasive Bladder Cancer Based On Immune Cell Infiltration

ObjectiveThe purpose of this study is to explore the key genes and signal pathways of muscle invasive bladder cancer and to construct a prognostic model of muscle invasive bladder cancer by analyzing the relevant data of TCGA and GEO database.Methods1.The data of 306 patients with bladder cancer(T2 or above)was downloaded from TCGA database as the experimental group.Downloading gene chip GSE13507 from GEO dataset and 62 cases of muscle invasive bladder cancer samples were included in the study as the verification group.2.The immune cells were quantified by CIBERSORT algorithm and ssGSEA,and the survival analysis and COX single factor correlation analysis of immune cells were carried out.3.The genes related to the target immune cells were screened and used as the input gene set for WGCNA analysis to screen the modules related to death and tumor recurrence4.The 30 genes with the highest connectivity in the module were selected for LASSO-Cox regression analysis,so as to build a risk value prediction model and draw a nomogram combined with other clinical traits5.After enrichment analysis of genes in the module,survival analysis of the gene set of the model was carried out to screen potential target genes,and the mutation sites and drug sensitivity information of potential target genes were analyzed by cBioPortal database and GDSC database.Result1.From the analysis of immune cell infiltration,it is concluded that eosinophils is a protective factor for patients with muscle invasive bladder cancer2.The green module associated with death(cor=0.15,p=0.008)and tumor recurrence(cor=0.19,p<0.001)was obtained by weighted gene co-expression network analysis of2689 genes with high correlation with eosinophils.3.The LASSO-Cox regression analysis was carried out of the 30 genes with the highest connectivity in the green module,and a risk value prediction model composed of 14 genes was constructed,and the age,sex,tumor grade,TNM stage and risk value model were integrated into the nomogram.The C-index value is 0.744,the correction curve and DCA curve also verify the prediction accuracy of the nomogram.4.Enrichment analysis of 227 genes in the green module showed that these genes were mainly enriched in spindles,chromosomes,intermediates,centromere,etc,and participated in the regulation of mitosis,chromosome segregation and mitosis through tubulin binding,adenosine triphosphatase activity,microtubule binding,kinesin driving activity and so on.They are also related to the following signal pathways:cell cycle,oocyte meiosis,cell senescence,transforming growth factor-β signal pathway,progesterone-mediated oocyte maturation and so on.5.Three potential target genes or drug targets were screened by survival analysis and oncomine online database verification:ASPM,KIF23 and CDC20.The mutation frequency of ASPM is 5%,the mutation frequency of KIF23 is 2.4%,the mutation frequency of CDC20 is 1%,and all of them are missense mutations.The ASPM mutation is mainly located in the calmodulin binding motif of IQ,and the KIF23 mutation is mainly located in the motor domain of the kinesin.Through GDSC database,we found ASPM mutation sensitive drugs:Nutlin-3a(-),Leflunomide.ConclusionWe constructed a prognostic model of muscle invasive bladder cancer based on immune cell infiltration,WGCNA and LASSO-Cox regression,which combines 14 genes and clinical characteristics such as age and sex.We also confirmed that the model is accurate and intuitive.The high expression of ASPM,KIF23 and CDC20 is related to the poor prognosis of muscle invasive bladder cancer,which may become new molecular markers or therapeutic targets.ASPM-sensitive drugs like Nutlin-3a and Leflunomide may provide a new direction for targeted therapy of muscle invasive bladder cancer.