Exploration of Potential Biomarker for Poor Prognosis in Peritoneal Dialysis-Related Peritonitis Based on Proteomics

Background: Peritoneal dialysis (PD) related peritonitis was the most critically important outcome in PD population. It not only had high incidence, but showed poor prognosis. There is an urgent need for predictive biomarkers of poor prognosis. Methods: Poor group was defined as peritonitis-associated catheter removal, hemodialysis transfer, and death, and good group was maintenance PD. Untreated overnight PD effluent from the proteomics cohort were detected using LC-MS/MS. The differential expression proteins (DEPs) were identified by Student’s t-test and Benjamini-Hochberg (BH) FDR correction. Optimal prognostic risk model were screened with 15 machine learning algorithms, and validated by 5-fold cross-validation. The potential biomarker was validated by ELISA. Results: The clinical characteristics were comparable between good and poor groups. A total of 254 proteins were identified in the dialysis effluent, and there were 17 DEPs (Student’s t test, FDR < 0.05, and [log2FC] > 0.75) (Figure 2 A-C). In the GO and KEGG analysis, DEPs were significantly enriched in maintaining vascular homeostasis and antimicrobial defense (Figure 2 D-E). The top 12 optimal machine learning models’ accuracies, recalls, and F1-scores all reached 1, which were constructed by ENR, SVM, and lasso. Based on DEPs and the predictive models, DEFA3 was the most potential biomarker for poor outcomes, and its levels were significantly increased in poor group [proteomics cohort: 25.65 (7.07, 31.20) ng/ml in poor group vs. 5.52 (3.04, 9.75) ng/ml in good group, P = 0.025; expanded cohort: 27.09 (4.56, 31.57) ng/ml vs. 6.28 (4.82, 8.27) ng/ml, P < 0.001]. Conclusion: DEFA3 might be a novel biomarker for peritonitis associated poor prognosis. Funding: Government Support – Non-U.S.Figure 1. The schematic workflow.Figure 2. DEFs between peritonitis with poor and good outcomes.