Screening of Genes That May be Associated with Gastric Cancer Using Bioinformatics and Artificial Intelligence Methods and Interpretation of IndividualBased Results

Aim: There is no screening technique for gastric cancer, which is one of the most common cancer types among the causes of death from cancer, and the early diagnosis of the disease is very low. Therefore, there is a need for early detection of the disease. The aim of this study is to identify potential genes that may be associated with gastric cancer by bioinformatics methods using open access gene expression data obtained from human gastric tumor tissues and normal gastric tissues, and also to classify the data with random forest (RF), one of the machine learning models, and to evaluate the genes that may be associated with the disease on an individual basis using LIME, one of the explainable artificial intelligence (XAI) models. Methods: Bioinformatics analyses of the data were performed using the limma package in the R programming language. In the modelling phase, classification was performed using RF model and the classification performance was evaluated with accuracy, balanced accuracy, sensitivity, specificity, positive predictive value, negative predictive value, and F1-score metrics. LIME, one of the XAI methods, was used to explain the applied model. Results: According to the results of bioinformatic analysis 3283 expression was found with statistically significant differences in gene expression levels between the two groups. As a result of modelling with RF the performance metrics obtained from the model were accuracy (96.7%), balanced accuracy (96.7%), sensitivity (93.3%), specificity (100%), positive predictive value (100%), negative predictive value (93.8%), and F1-score (96.6%), respectively. According to the results of XAI model, CORO1C, CAPN13, SST, GGT6, ARSD, CYP3A5 genes were found to be effective in tumor formation. Conclusion: Genes that may be associated with gastric cancer were identified by bioinformatics and machine learning models. Based on the changes of the identified genes in individuals, future studies can be directed or individual trials can be conducted for the treatment of the disease.