Using two-sample Mendelian randomization to study the causal relationship between BMI and pancreatic cancer and applying bioinformatics methods to identify important genes suggestive of poor prognosis in pancreatic cancer

Abstract Background： A causal relationship between body mass index (BMI) and pancreatic cancer (PC) has been reported; however, due to traditional observational studies, residual confounding, reverse causation, and the presence of measurement error, a causal relationship between them has not yet been confirmed. The present study used Mendelian randomization (MR) to investigate the causal relationship between BMI and PC. Since pancreatic cancer is usually detected at an advanced stage most treatment options are ineffective. Therefore, this study used a bioinformatics approach to explore prognostic biomarkers of PC as a target to improve therapeutic efficacy and better understand the underlying mechanisms. Methods: We conducted a two-sample Mendelian randomization (MR) study to assess their causal relationship. Three MR analysis methods were used for causal inference and inverse variance weighting (IVW) was chosen as the primary method. Mendelian randomized polytomous residuals and outliers (MR-PRESSO) were used to exclude outlier SNPs. To assess the robustness of the MR results, a "leave-one-out" analysis was performed. In addition, to identify important genes suggestive of poor prognosis in PC. We first selected GSE71989, GSE32676, and GSE16515 from the Gene Expression Omnibus (GEO) system. Second, we applied the GEO2R online tool and Venn diagram software to obtain the common differentially expressed genes (DEGs) in the above three datasets. Third, these databases were analyzed using the Database for Annotation, Visualization, and Integrated Discovery (DAVID), including molecular function (MF), cellular component (CC), biological process (BP), and KEGG pathways. Fourth, we established a protein-protein interaction (PPI) network and then applied cell-based molecular complex detection (MCODE) to further analyze the DEGs to identify some core genes. In addition, these core EDGs were imported into the unlcan online database for important prognostic information (P < 0.05). In addition, we further verified the expression of DEGs between PC tissues and normal pancreatic tissues by gene expression profiling interaction analysis (P < 0.05) to obtain qualified DEGs. We then reanalyzed the saturation of KEGG pathway enrichment of these DEGs. Finally, important genes suggestive of poor prognosis in PC were obtained. Results: Our MR study provided evidence that no heterogeneous genes were identified, no horizontal pleiotropy was detected by MR-PRESSO analysis, and there was a causal relationship between BMI and PC (IVW, OR = 1.26, 95% CI = [1.012, 1.572], P = 0.0392). Sensitivity analysis confirmed the robustness of the MR results. In addition, bioinformatics analysis identified CDK1 and RRM2 as new effective targets to improve the prognosis of PC patients. A large number of studies have demonstrated that these two genes are associated with the progression of different types of cancer, but few studies have been reported on these two genes in PC. Therefore, the data from our study may provide useful information and direction for future research in PC. Conclusion: Our study confirms the causal relationship between BMI and PC and identifies CDK1, and RRM2 genes as possible key players in the progression of PC. The present study may provide insights to explore the underlying mechanisms of PC development and prevention.