Diagnosis and prognosis prediction model for digestive system tumors based on immunologic gene sets

Abstract

According to 2020 global cancer statistics, digestive system tumors (DST) are ranked first in both incidence and mortality. This study systematically investigated the immunologic gene set (IGS) to discover effective diagnostic and prognostic biomarkers. Gene set variation (GSVA) analysis was used to calculate enrichment scores for 4,872 IGSs in patients with digestive system tumors. Using the machine learning algorithm XGBoost to build a classifier that distinguishes between normal samples and cancer samples, it shows high specificity and sensitivity on both the validation set and the overall dataset (area under the receptor operating characteristic curve [AUC]: validation set = 0.993, overall dataset = 0.999). IGS-based digestive system tumor subtypes (IGTS) were constructed using a consistent clustering approach. A risk prediction model was developed using the Least Absolute Shrinkage and Selection Operator (LASSO) method. DST is divided into three subtypes: subtype 1 has the best prognosis, subtype 3 is the second, and subtype 2 is the worst. The prognosis model constructed using nine gene sets can effectively predict prognosis. Prognostic models were significantly associated with tumor mutational burden (TMB), tumor immune microenvironment (TIME), immune checkpoints, and somatic mutations. A composite nomogram was constructed based on the risk score and the patient’s clinical information, with a well-fitted calibration curve (AUC = 0.762). We further confirmed the reliability and validity of the diagnostic and prognostic models using other cohorts from the Gene Expression Omnibus database. We identified diagnostic and prognostic models based on IGS that provide a strong basis for early diagnosis and effective treatment of digestive system tumors.