Screening amino acid metabolism-related gene signature for recurrence prediction of colon adenocarcinoma

Abstract

Objective: This study aimed to develop an amino acid metabolism-related gene signature for recur-rence prediction in colon adenocarcinoma (COAD).Methods: We downloaded RNA sequencing profiles of COAD from The Cancer Genome Atlas(TCGA) as a training set and GSE39582 from the Gene Expression Omnibus database as a validationset. Differentially expressed RNAs (DERs) were screened from recurrence tumor samples by definedthresholds. The amino acid metabolism-related gene signature was identified by LASSO cox regressionanalysis. Independent prognostic clinical factors were also assessed by using survival analysis. A riskscore (RS) signature was established and validated in two independent datasets.Results: We obtained 498 differentially expressed mRNAs and 71 differentially expressed lncRNAsbased on data mining. Compared with amino acid metabolism genes of Gene Set Enrichment Analysis da-tabase, we screened 197 overlapped DERs. A twelve genes-based RS signature was established. This modelexhibited a high accuracy for recurrence prediction with an area under the ROC curve (AUC) of 0.924and 0.843 in TCGA and GSE39582, respectively. In addition, pathological stage and RS model status wereidentified as independent clinical factors associated with recurrence. The combined model integratingthese two factors reached a higher AUC value of 0.940 in the TCGA dataset and 0.876 in the validation set.Conclusion: We established a high accuracy prognostic model for recurrence prediction. Our find-ings suggested that the combined model can identify high-risk recurrence COAD patients and might bea reliable tool for decision-making in a clinic