Prognostic gene biomarker identification in liver cancer by data mining

Liver cancer is a common cancer that enormously threatens the health of people worldwide. With the continuous advances of high-throughput gene sequencing technology and computer data mining technology, researchers can understand liver cancer based on the current accumulation of gene expression data and clinical information.

We downloaded the TCGA data of liver cancer on the cancer-related website (https://genome-cancer.ucsc.edu/proj/site/hgHeatmap/), comprising 438 patients and 20,530 genes. After removing some patients with missing survival data, we collected 397 patients' samples. Our data were collected from a public database without real patient participation. While matching the patient samples in the gene expression spectrum, we attained 330 samples with primary tumors and 50 samples with normal solid tissue.

After the 330 tumor tissue samples were randomized into two equal-numbered groups (one is a training set, and the other is a test set), we selected 26 gene biomarkers from the training set and validated them in the test set. Based on the selected 26 gene biomarkers, RBM14, ALG11, MAG, SETD3, HOXD10 and other 26 genes were considered independent risk factors for the prognosis of liver cancer, and genes such as GHR significantly affect human growth hormone for liver cancer. The findings discovered that low-risk patients survived remarkably better than the high-risk patients (P<0.001), and the area under the curve (AUC) of receiver operating characteristic curve (ROC) was greater than 0.5.

Our numerical results showed that these 26 gene biomarkers can be used to guide the effective prognostic therapy of patients with liver cancer.