Comprehensive analysis of the prognosis of gastric cancer on the basis of three disulfidptosis-related lncRNA signatures

Abstract This study aimed to investigate the relationship between disulfidptosis-related long non-coding RNAs (DRLs) and gastric cancer (GC), as well as their association with GC prognosis, providing new insights and potential therapeutic targets for prognostic evaluation and personalized treatment of GC. We retrieved gene expression, clinical, mutation, and copy number data associated with GC from The Cancer Genome Atlas (TCGA) database and processed and analyzed the data using R software (version 4.2.2) and Perl software (version 5.30.0). Utilizing univariate Cox regression analyses (with Benjamini-Hochberg false discovery rate [FDR] correction), Lasso regression analysis (10-fold cross-validation), and multivariate Cox regression analyses, we established three prognostic prediction models for GC through a rigorous multi-screening strategy. The samples were randomly divided into training and testing sets (repeated three times with different subsets), and 200 initial models were generated. Three optimal signatures were ultimately selected based on predefined objective criteria, including concordance index (C-index) > 0.6, time-dependent area under the ROC curve (AUC) > 0.6 at 1, 3, and 5 years, statistically significant survival differences in both training and testing cohorts ( P < 0.05), and principal component analysis (PCA) demonstrating clear risk-group separation. Additionally, differential analysis, Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG), and Gene Set Enrichment Analysis (GSEA) were performed to explore the potential functions and pathways of these lncRNAs in GC. Immune correlation analysis was conducted to assess the relationship between these lncRNAs and the tumor microenvironment. Drug sensitivity analysis was performed using oncoPredict to identify in silico drug response predictions. We successfully developed three independent disulfidptosis-related GC prognostic models through a multi-screening strategy. These models significantly predicted the prognosis of GC patients and were associated with clinical parameters such as age, grade, and stage. Direct comparison revealed that the three models exhibited comparable predictive accuracy, with Signature 3 showing marginally superior C-index values in certain subgroups. Differential analysis, GO, KEGG, and GSEA revealed potential functions and pathways associated with disulfidptosis-related lncRNAs in GC, including focal adhesion and ECM-receptor interaction. Immune correlation analysis indicated that these lncRNAs were associated with tumor mutation burden (TMB), M2 macrophages, and mast cells, among other immune-related factors. In silico drug sensitivity analysis identified seven drugs (AZD6738, BMS-345541, dabrafenib, MK-1775, ML323, oxaliplatin, and savolitinib) with predicted differential sensitivity between risk groups. This study established three independent multi-screening disulfidptosis-related GC prognostic models, offering complementary indicators for GC prognosis evaluation. The three-signature strategy enhances robustness by reducing the risk of single-model overfitting. Immune analysis suggests that TMB, M2 macrophages, and mast cells may be involved in the GC immune microenvironment. The focal adhesion pathway may be associated with GC progression. Furthermore, we identified potential drugs with predicted sensitivity to GC, providing in silico candidates for personalized treatment strategies. However, all findings are computational and require external validation and experimental confirmation.