Development Of Prognostic Model Of Colorectal Cancer Based On LncRNAs With Different Features And Bioinformatics Analysis

Objective:This study aimed to construct the prognostic risk score model of colorectal cancer(CRC)based on different feature selection methods and explore the potential prognostic value of cuproptosis-related Lnc RNAs(CRLs)in CRC,provide new therapeutic targets for CRC patients,and contribute to the development of clinical treatment.Methods:First,we downloaded RNA-Seq data and clinical information of CRC patients from the TCGA database.The CRLs were obtained based on Pearson correlation analysis and identified the potential subtype using the "Consensus Cluster Plus" package.Then,six feature selection methods processing the survival censored data(Lasso-Cox and RSF)and survival classification data(XGBoost,RF,Light GBM,and GBDT)were separately used to select the differential Lnc RNAs between subtypes and construct risk score models.Besides,we compare the prediction performance for the six models and select the best risk score model.Third,we divided CRC patients into high-risk and low-risk groups according to the median risk score(predictive value),performed survival analysis,and constructed a nomogram model.Moreover,the GEO dataset was used for external verification.The Kaplan-Meier method,ROC curves,and decision clinical analysis curve(DCA)were used to evaluate the predictive ability of models.Finally,we used gene function analysis(GO)and signaling pathway enrichment analysis(KEGG)to enrich the differential genes between the high-risk and low-risk groups.Furthermore,we used the CIBERSORT and ESTIMATE algorithms to assess the difference in immune cell infiltration between the high-risk and low-risk groups and calculated the correlation between risk score and drug sensitivity using Spearman correlation analysis.Results:1.A total of 1831 cuproptosis-related genes(CRGs)were identified.There are different survival outcomes between the two potential subtypes groups.Finally,73 CRLs were obtained according to the intersection of differential genes of subgroups with the GEO dataset.2.The feature selection results showed that the Lasso-Cox risk score model had the best predictive performance.3.CRC patients were divided into the high-risk and the low-risk group by Lasso-Cox.The results of the Log-rank test showed that there are significant survival differences between the high-risk group and the low-risk group.The ROC curve showed that the AUCs of 1-,3-and 5-year survival in the TCGA dataset were 0.639,0.649,and 0.660,respectively.The AUCs of the ROC were 0.625,0.654,and 0.611 for 1-,3-and 5-year survival in the GEO dataset,respectively.The result of the multivariable Cox model contained clinical information showing that the risk score could be an independent CRC prognostic factor independent of other clinical characteristics.In addition,we constructed a nomogram model integrating clinical characteristics and the risk score to predict the survival rate of CRC patients at 1,3,and 5 years.The ROC curves of 1-,3-and 5-year survival rates based on the training set were 0.799,0.804,and 0.831,respectively.Based on the testing set,the ROC curves of 1-,3-and 5-year survival rates were 0.738,0.690,and 0.685,respectively.4.GO and KEGG analyses revealed that the differential genes in the high-risk and low-risk groups were significantly enriched in humoral immunity,antibacterial humoral immunity,and IL 17 signaling pathways.The result of immune infiltration showed that there are significant differences in B cells native,T cells CD4+memory activated,T cells follicular helper,T cells regulatory(Tregs),Macrophages M1,Dendritic cells resting,Mast cells resting,Mast cells activated and Neutrophils between the high-risk group and the low-risk group.5.The results of drug sensitivity analysis showed that the high-risk group had high sensitivity in Camptothecin,Darafenib,Elephantin,Erlotinib,Foretinib,Irinotecan,and Oxaliplatin,while the low-risk group was sensitive to Carmustine and Bcl-2inhibitor Navitoclax.Conclusions:1.The Lasso-Cox method has a significant advantage in CRLs screening,followed by the RSF-VIMP method,and machine learning was weak.2.A risk score model containing three CRLs(LINC01138,ALMS1-IT1,and LINC01410)was constructed,which could classify patients into the high-risk group and the low-risk group.There are significant differences in survival outcomes and immune function between the high-risk and low-risk groups.In addition,this risk score can be used as an independent prognostic factor of the prognosis of CRC.3.LINC01138,ALMS1-IT1,and LINC01410 may be potential prognostic molecular biomarkers and therapeutic targets for CRC.