Gene expression and machine learning techniques uncover corneal biomarkers associated with oxidative stress in the myopia progression.

This study aimed to identify and validate corneal biomarkers underlying myopia using human corneal tissues. Gene expression data were analyzed to investigate myopia development, where candidate genes were selected by identifying differentially expressed genes and intersecting them with oxidative stress-related genes. Machine learning techniques were employed to identify key biomarkers, and a nomogram was constructed to predict myopia risk. Utilizing corneal stromal tissues from patients who undergoing Small Incision Lenticule Extraction (SMILE) surgery, the expression levels of ATF3, GRIN2B, and GSTM3 were found to be significantly lower in the high myopia group (≤ -6.00 D) compared to the low myopia group (≥ -3.00 D and < 0 D). These biomarkers were also found to be closely associated with differential immune cell infiltration, particularly involving CD8 + T cells and eosinophils. A diagnostic nomogram was developed and showed strong discriminative potential in the discovery set. However, its predictive performance and clinical utility should be further validated in independent cohorts. ATF3, GRIN2B, and GSTM3 have emerged as promising oxidative stress-related biomarkers with significant potential for understanding myopia pathology.