Development and clinical validation of a novel 9-gene prognostic model based on multi-omics in pancreatic adenocarcinoma

The prognoses of patients with pancreatic adenocarcinoma (PAAD) remain poor due to the lack of biomarkers for early diagnosis and effective prognosis prediction. RNA sequencing, single nucleotide polymorphism, and copy number variation data were downloaded from The Cancer Genome Atlas (TCGA). Univariate Cox regression was used to identify prognosis-related genes. GISTIC 2.0 was used to identify significantly amplified or deleted genes, and Mutsig 2.0 was used to analyze the mutation data. The Lasso method was used to construct a risk prediction model. The Rms package was used to evaluate the overall predictive performance of the signature. Finally, Western blot and polymerase chain reaction were performed to evaluate gene expression. A total of 54 candidate genes were obtained after integrating the genomic mutated genes and prognosis-related genes. The Lasso method was used to ascertain 9 characteristic genes, including UNC13B, TSPYL4, MICALI, KLHDC7B, KLHL32, AIME ARHGAP18, DCBLDI, and CACNA2D4. The 9-gene signature model was able to help stratify samples at risk in the training and external validation cohorts. In addition, the overall predictive performance of our model was found to be superior to that of other models. KLHDC7B, AIM1, DCBLD1, TSPYL4, and MICAL1 were significantly highly expressed in tumor tissues compared to normal tissues. ARHGAP18 and CACNA2D4 had no difference in expression between tumor and normal tissues. UNC13B and KLHL32 expression in the normal group was higher than in the tumor group. The 9-gene signature constructed in this study can be used as a novel prognostic marker to predict the survival of patients with pancreatic adenocarcinoma.

Automated summary

This study integrated genomic mutated genes and prognosis-related genes to construct a predictive model with 9 characteristic genes, which could help identify samples at risk and demonstrated superior overall predictive performance in training and external validation cohorts.