Identifying core genes in keloid and investigating immune infiltration and pan-cancer associations using eQTL and machine learning.

Motivation: Keloid is a fibroproliferative skin disorder characterized by excessive fibroblast proliferation and abnormal extracellular matrix accumulation. It manifests as continuous growth, redness, itching, and pain, with a high recurrence rate. The pathogenesis of keloid is complex, with genetics and gene mutations increasingly recognized as critical risk factors. This condition exhibits familial predisposition and clustering, with individuals of darker skin tones at greater risk. To elucidate the genetic factors underlying keloid development, this study integrates bioinformatics and Mendelian randomization (MR) approaches to identify core genes associated with keloid, providing novel insights into its pathogenesis, treatment, and prognosis. Results: Bioinformatics and Mendelian randomization analyses identified two intersecting genes, CCND2 and KLF4, as core genes associated with keloid. MR analysis revealed that CCND2 is causally associated with keloid [inverse variance weighted (IVW) odds ratio (OR): 1.410; 95% confidence interval (CI): 1.001-1.985, Availability and implementation: Firstly, bioinformatics analysis identified differentially expressed genes (DEGs) from the keloid GEO datasets. Secondly, MR was applied to eQTL and keloid GWAS datasets to identify candidate genes. Overlapping genes were derived by intersecting DEGs with MR candidate genes. Causal relationships between overlapping genes and keloids were analyzed using five MR methods, identifying core genes significantly associated with keloid pathogenesis. Cochran's Q test and MR-Egger intercept analysis evaluated heterogeneity and pleiotropy in MR results. GO, KEGG, and GSEA enrichment analyses were conducted to explore core gene functions. Finally, validation and TCGA pan-cancer analyses were conducted on the core genes.