Identification and validation of arsenic-associated genes and risk model for predicting lung cancer.

Abstract

e13631 Background: Lung cancer, the leading cause of cancer-related mortality globally, constitutes approximately 12.4% of all cancer diagnoses. Environmental factors, particularly tobacco smoking, account for 90% of bronchogenic carcinomas. Arsenic trioxide, a carcinogen in tobacco smoke, is linked to DNA damage and tumorigenesis, increasing lung cancer risk in exposed individuals, including smokers. Despite evidence associating arsenic with other cancers, such as bladder cancer, its relevance in lung cancer remains insufficiently studied. This research investigates arsenic-associated genes in lung cancer to identify predictive biomarkers and therapeutic targets. Methods: A dataset comprising 959 samples (842 lung carcinoma and 117 normal tissues) was curated from the Gene Expression Omnibus (GEO), including demographic and clinical data (e.g., age, smoking status, gender, cancer stage, and subtype). Data generated using the Affymetrix GeneChip Human Genome U133 Plus 2.0 Array platform underwent normalization via frozen robust multiarray analysis (fRMA). Principal component analysis (PCA) assessed variability between tumor and normal tissues. Differential expression analysis identified genes with significant fold changes, supported by statistical metrics (t-statistics, p-values, false discovery rates). Logistic regression modeling was performed, with predictive accuracy evaluated using receiver operating characteristic (ROC) curve analysis. Results: PCA revealed distinct separation between tumor and normal tissues (PC1: 10.1%; PC2: 7.8%). Among 147 arsenic-associated genes previously linked to bladder cancer, 88 genes exhibited significant differential expression in lung cancer. Functional analysis highlighted six genes ARHGEF10, ADARB1, SEC14L1, CBX7, GYPC, and CRIM1, as key players in tumor progression, influencing processes such as tumor signaling, RNA editing, lipid transport, chromatin remodeling, cellular adhesion, and angiogenesis. A predictive model incorporating four of these genes demonstrated strong performance, achieving an AUC of 0.886 (95% CI: 0.838-0.932) with all contributing significantly (p < 0.05). Conclusions: This study identifies four robust biomarkers among 88 significant genes for lung cancer risk prediction, achieving high model accuracy. These findings highlight the utility of arsenic-associated genes in assessing cancer risk and advancing understanding of heavy metal toxicity in lung cancer. Further investigation of these biomarkers could provide insights into their molecular roles and enhance predictive and therapeutic strategies.