2026-10-13T17:59:45Z https://api.figshare.com/v2/oai

oai:figshare.com:article/33975224 2026-09-23T17:26:53Z category_12 category_13 category_873 category_46 category_734 portal_5 item_type_3 month_year_09_2026

R script for SHAP-based interpretation analysis. Qingao Deng (25099004) Junjie Wang (172690) Xing Wang (154377) Lu Qi (160573) Cell Biology Genetics Chemical Sciences not elsewhere classified Immunology Biological Sciences not elsewhere classified ‑ cell reduction temporomandibular joint affected shapley additive explanations potential clinical links offers novel insights nf ‑ κb machine learning screening lipid atherosclerosis pathways immune infiltration assessment il ‑ 17 deoxycholic acid –< chronic inflammatory diseases adaboost ensemble identified therapeutic target validation th17 cells drive resting mast cells shared pathogenic genes potential therapeutic targets div >< p therapeutic targets plasma cells shared mechanisms core targets vcan </ study aimed specific binding significantly enriched selp </ rheumatoid arthritis molecular docking mol ), model interpretability mmp1 </ mmp </ may serve lgalsl </ itgam </ integrative bioinformatics immunological characteristics geo database fam46c </ elovl4 </ cxcl1 </ <div>Periodontitis (PD) and systemic juvenile idiopathic arthritis (sJIA) are chronic inflammatory diseases with potential clinical links, yet their shared molecular mechanisms remain unclear—patients with JIA face elevated periodontal risk, the temporomandibular joint affected in JIA shares inflammatory pathways with PD, Th17 cells drive both conditions, and glucocorticoid therapy may exacerbate periodontal vulnerability. This study aimed to explore the shared mechanisms and potential therapeutic targets of PD and sJIA using genetic expression data from the GEO database and performed comprehensive bioinformatics analyses, including differential expression gene analysis, weighted gene co‑expression network analysis (WGCNA), functional enrichment analysis, protein‑protein interaction network construction, and machine learning across 175 predictive models with the Ridge + AdaBoost ensemble identified as the best‑performing combination, followed by SHapley Additive exPlanations (SHAP) for model interpretability, CIBERSORT for immune infiltration assessment, and molecular docking with 100‑ns molecular dynamics simulations for therapeutic target validation. We identified 37 shared candidate genes between PD and sJIA, which were significantly enriched in IL‑17, NF‑κB, rheumatoid arthritis, and lipid atherosclerosis pathways, and machine learning screening further selected six core diagnostic genes (FAM46C, CXCL1, SELP, LGALSL, ELOVL4, VCAN), with SELP demonstrating the most robust cross‑model, cross‑dataset diagnostic value (AUC > 0.8); SHAP analysis confirmed SELP as a stable risk‑driving predictor across all models, while CXCL1 consistently showed protective effects. Immune infiltration revealed shared neutrophil elevation and CD8 ⁺ T‑cell reduction, with conserved CXCL1 and SELP correlations with neutrophils and resting mast cells, and FAM46C with plasma cells.Drug‑target network analysis identified IL1B, MMP1, and ITGAM as core targets, and molecular docking yielded strong binding affinities for deoxycholic acid–MMP1 (−7.6 kcal/mol), kaempferol–IL1B (−7.2 kcal/mol), kaempferol–ITGAM (−6.9 kcal/mol), with MD simulations confirming stable and specific binding. Collectively, this study explores shared genetic and immunological characteristics between PD and sJIA, suggests that SELP may serve as a critical cross‑disease diagnostic biomarker, and offers novel insights into their pathogenesis and potential therapeutic targets that warrant further experimental validation.</div> 2026-09-23T17:26:43Z Dataset Dataset 10.1371/journal.pcbi.1014669.s008 https://figshare.com/articles/dataset/_p_R_script_for_SHAP-based_interpretation_analysis_p_/33975224 CC BY 4.0