A novel risk model based on the correlation between the expression of basement membrane genes and immune infiltration to predict the invasiveness of pituitary adenomas

Abstract

Objective: Invasive pituitary adenomas (IPAs) are common tumors of the nervous system tumors for which invasive growth can lead to difficult total resection and a high recurrence rate. The basement membrane (BM) is a special type of extracellular matrix and plays an important role in the invasion of pituitary adenomas (PAs). The aim of this study was to develop a risk model for predicting the invasiveness of PAs by analyzing the correlation between the expression of BM genes and immune infiltration. Methods: Four datasets, featuring samples IPAs and non-invasive pituitary adenomas (NIPAs), were obtained from the Gene Expression Omnibus database (GEO). R software was then used to identify differentially expressed genes (DEGs) and analyze their functional enrichment. Protein-protein interaction (PPI) network was used to screen BM genes, which were analyzed for immune infiltration; this led to the generation of a risk model based on the correlation between the expression of BM genes and immunity. A calibration curve and receiver operating characteristic (ROC) curve were used to evaluate and validate the model. Subsequently, the differential expression levels of BM genes between IPA and NIPA samples collected in surgery were verified by Quantitative Polymerase Chain Reaction (qPCR) and the prediction model was further evaluated. Finally, based on our analysis, we recommend potential drug targets for the treatment of IPAs. Results: The merged dataset identified 248 DEGs that were mainly enriching in signal transduction, the extracellular matrix and channel activity. The PPI network identified 11 BM genes from the DEGs: SPARCL1, GPC3, LAMA1, SDC4, GPC4, ADAMTS8, LAMA2, LAMC3, SMOC1, LUM and THBS2. Based on the complex correlation between these 11 genes and immune infiltration, a risk model was established to predict PAs invasiveness. Calibration curve and ROC curve analysis (area under the curve [AUC]: 0.7886194) confirmed the good predictive ability of the model. The consistency between the qPCR results and the bioinformatics results confirmed the reliability of data mining. Conclusion: Using a variety of bioinformatics methods, we developed a novel risk model to predict the probability of PAs invasion based on the correlation between 11 BM genes and immune infiltration. These findings may facilitate closer surveillance and early diagnosis to prevent or treat IPAs in patients and improve the clinical awareness of patients at high risk of IPAs.