Exploration of endoplasmic reticulum stress-related gene markers in amyotrophic lateral sclerosis: a comprehensive analysis of bioinformatics and machine learning.

This study aimed to investigate potential biomarkers related to Endoplasmic reticulum (ER) stress in Amyotrophic lateral sclerosis (ALS) through a comprehensive bioinformatic approach. The gene expression profiles of ALS patients and healthy controls were downloaded from the Gene Expression Omnibus (GEO) database. ER stress-related genes were collected from the MSigDB databases and document literature. The "limma" R package was employed to detect the differentially expressed ER stress-related genes (DE-ERSGs). Three methods of machine learning were applied to select the hub DE-ERSGs. ROC curves were conducted to evaluate model performance. An external dataset was chosen to evaluate the diagnostic capability of hub genes. The CIBERSORT algorithm was used to evaluate the immune cell infiltration characteristics. Additionally, we constructed a systematic ceRNA regulatory network using Cytoscape software and predicted the possible drug candidates using the Enrichr platform. Molecular docking analysis was used to further validate the binding ability of the candidate drug molecules to the hub genes. Six hub DE-ERSGs (ABCA1, CKAP4, TOR1AIP1, MMP9, EDC4, and ALPP) were identified, and the related models performed well. These hub genes were concentrated in multiple pathways and related to various immune cells. Drugs such as nitroglycerin, diazepam, FENRETINIDE, and edaravone exhibited good binding affinity to the hub genes, indicating that they may be promising drugs for the management of ALS. This study revealed the essential role of ER stress in the pathogenesis of ALS from an integrative perspective, providing guidance for the development of new therapeutic targets and diagnostic strategies.