Development of a blood proteins-based model for bronchopulmonary dysplasia prediction in premature infants.

BACKGROUND: Bronchopulmonary dysplasia (BPD) is the most common chronic pulmonary disease in premature infants. Blood proteins may be early predictors of the development of this disease. METHODS: In this study, protein expression profiles (blood samples during their first week of life) and clinical data of the GSE121097 was downloaded from the Gene Expression Omnibus. Weighted gene co-expression network analysis (WGCNA) and differential protein analysis were carried out for variable dimensionality reduction and feature selection. Least absolute shrinkage and selection operator (LASSO) were conducted for BPD prediction model development. The performance of the model was evaluated by the receiver operating characteristic (ROC) curve, calibration curve, and decision curve. RESULTS: The results showed that black module, magenta module and turquoise module, which included 270 proteins, were significantly correlated with the occurrence of BPD. 59 proteins overlapped between differential analysis results and above three modules. These proteins were significantly enriched in 253 GO terms and 11 KEGG signaling pathways. Then, 59 proteins were reduced to 8 proteins by LASSO analysis in the training cohort. The proteins model showed good BPD predictive performance, with an AUC of 1.00 (95% CI 0.99-1.00) and 0.96 (95% CI 0.90-1.00) in training cohort and test cohort, respectively. CONCLUSION: Our study established a reliable blood-protein based model for early prediction of BPD in premature infants. This may help elucidate pathways to target in lessening the burden or severity of BPD.