Prognostic assessment capability of a five-gene signature in pancreatic cancer: a machine learning based-study.

Background: A prognostic assessment method with good sensitivity and specificity plays an important role in the treatment of pancreatic cancer patients. Finding a way to evaluate the prognosis of pancreatic cancer is of great significance for the treatment of pancreatic cancer.

Methods: In this study, GTEx dataset and TCGA dataset were merged together for differential gene expression analysis. Univariate Cox regression and Lasso regression were used to screen variables in the TCGA dataset. Screening the optimal prognostic assessment model is then performed by gaussian finite mixture model. Receiver operating characteristic (ROC) curves were used as an indicator to assess the predictive ability of the prognostic model, the validation process was performed on the GEO datasets.

Results: Gaussian finite mixture model was then used to build 5-gene signature (ANKRD22, ARNTL2, DSG3, KRT7, PRSS3). Receiver operating characteristic (ROC) curves suggested the 5-gene signature performed well on both the training and validation datasets.

Conclusions: This 5-gene signature performed well on both our chosen training dataset and validation dataset and provided a new way to predict the prognosis of pancreatic cancer patients.