Development and validation of a four-lipid metabolism gene signature for diagnosis of pancreatic cancer.

Abnormal lipid metabolism is closely related to the malignant biological behavior of tumor cells. Such abnormal lipid metabolism provides energy for rapid proliferation, and certain genes related to lipid metabolism encode important components of tumor signaling pathways. In this study, we analyzed pancreatic cancer datasets from The Cancer Genome Atlas (TCGA) and searched for prognostic genes related to lipid metabolism in the Molecular Signature Database. A risk score model was built and verified using the GSE57495 dataset and International Cancer Genome Consortium (ICGC) dataset. Four molecular subtypes and 4,249 differentially expressed genes were identified. The differentially expressed genes obtained by Weighted Gene Co-expression Network Construction (WGCNA) analysis were intersected with 4249 differentially expressed genes to obtain a total of 1340 differentially expressed genes. The final prognosis model included CA8, CEP55, GNB3 and SGSM2, and these had a significant effect on overall survival (OS). The area under the curve (AUC) at 1 years, 3 years and 5 years was 0.72, 0.79 and 0.87, respectively. These same results were obtained using the validation cohort. Survival analysis data showed that the model could stratify the prognosis of patients with different clinical characteristics and the model has clinical independence. Functional analysis indicated that the model is associated with multiple cancer-related pathways. Compared with published models, our model has a higher C-index and greater risk value. In summary, this 4-gene signature is an independent risk factor of pancreatic cancer survival, and may be an effective prognostic indicator.