Identification and validation of biomarkers related to integrated stress response in ischemic stroke based on transcriptomics

Ischemic stroke (IS) is a prevalent cerebrovascular disorder. Integrated stress response (ISR) plays an important role in multiple cerebrovascular diseases, its specific involvement and associated biomarkers in IS remain incompletely understood. This study aimed to identify and validate ISR-related gene (ISR-RG) biomarkers for IS.

Transcriptomic datasets related to IS were obtained from public databases, and a curated list of ISR-RGs was derived from the literature. Differentially expressed genes (DEGs) between IS patients and healthy controls were identified through differential expression analysis of the GSE16561 dataset. Candidate genes were then obtained by intersecting DEGs with ISR-RGs.

Subsequently, potential biomarkers were pinpointed using machine learning algorithms (LASSO and SVM-RFE), validated by receiver operating characteristic curve analysis, and confirmed by expression consistency across datasets. Correlation analysis among the final biomarkers was performed. A predictive nomogram was developed and its performance assessed via calibration and decision curve analyses.

Furthermore, gene set enrichment analysis (GSEA), drug-target prediction, and RNA N6-methyladenosine (m6A) methylation site prediction were conducted to elucidate the underlying molecular mechanisms. Finally, the expression levels of the identified biomarkers were experimentally validated in clinical samples using reverse transcription quantitative PCR (RT-qPCR). Our analysis identified 600 DEGs in the GSE16561 cohort.

BNIP3L, G6PD and TAOK1 genes were established as ISR-related biomarkers by integration method. We combined the three biomarkers to analyze the nomogram, and the results showed good prediction accuracy. GSEA showed that all three biomarkers were co-enriched in purine metabolic pathway.

Drug predictive analysis identified 43 potential therapeutic compounds for these biomarkers, including imatinib, sodium nitrate and ofloxacin. The prediction of m6A methylation sites showed that TAOK1 had multiple modification sites, and G6PD and BNIP3L had fewer but different modification spectra. Finally, we used RT-qPCR to verify the above markers in clinical samples, which was consistent with the bioinformatics prediction.