Exploring the relationship between graft dysfunction with serum metabolites and inflammatory proteins: integrating Mendelian randomization, single-cell analysis, machine learning, and SHAP methods for comprehensive analysis

Abstract

Background The exact mechanism of graft dysfunction has not been fully clarified. We aimed to explore the causal effects of serum metabolites on graft dysfunction and the mediating role of inflammatory proteins.Methods We first analyzed the causal association of 1,091 serum metabolites and 91 inflammatory proteins with graft dysfunction by Mendelian randomization (MR) to find the most important indicators and calculated the mediating role of inflammatory proteins. Second, we further intergrated the single-cell analysis, machine learning, and Shapley Additive exPlanations (SHAP) methods to validate the role of the inflammatory protein in rejected transplanted kidneys.Results In MR analysis, the inverse variance weighted (IVW) method showed that N-acetylaspartyl glutamate (NAAG) levels reduced the risk of graft dysfunction (OR 0.648, 95% CI 0.522–0.805). Interleukin-2 receptor subunit beta (IL-2RB) mediated the association between NAAG and graft dysfunction with a mediation ratio of 11.35% (95% CI 1.85%–24.69%). In transplanted kidneys, IL-2RB is predominantly expressed elevated in T cells from rejection samples and is associated with decreased renal function. 10 hub genes were identified based on T cells with high and low expression of IL-2RB, of which NGboost reached AUC 0.77. The SHAP method obtained the gene with the greatest model contribution: the IL-2RB gene.Conclusion NAAG plays a protective role in graft dysfunction, and part of the effect is mediated by inhibition of IL-2RB, which is verified as an important gene that functions in transplant kidney rejection. NAAG and IL-2RB may be novel molecular markers and therapeutic targets for graft dysfunction.

Keywords

• N-acetylaspartylgluta­mate
• graft dysfunction
• IL-2RB
• Mendelian randomization
• machine learning
• SHAP