Аңдатпа
Background and objectives: Bronchopulmonary dysplasia (BPD) remains a significant complication linked to prematurity, especially in infants with very and extremely low birth weights. A major challenge in identifying prognostic biomarkers is the biological noise in blood samples, which masks the specific signals of BPD. This study aimed to identify candidate miRNA markers in the blood and develop integrated models for early BPD risk assessment. Materials and methods: A prospective cohort study was conducted on preterm infants with birth weights below 1,500 g. Among the 41 infants in the cohort, 20 developed BPD and 21 did not. Peripheral blood samples were collected at 10-14 days of life, after which samples were analyzed using GeneChip™ miRNA 4.1. Differential expression was determined using limma, with a false discovery rate threshold of 0.05. To identify candidate miRNAs associated with systemic inflammatory markers and exclude non-specific signals from modeling, Spearman's correlation analysis was subsequently performed. Next, a predictive model was developed using LASSO regression and validated through leave-one-out cross-validation. All statistical analyses were performed using R. Results: < 0.05). Spearman correlation analysis revealed no significant associations between the five candidate miRNAs and systemic inflammatory parameters. Four miRNAs were selected based on the highest composite evidence scores across statistical significance, discriminative performance, and penalized regression stability criteria, comprising hsa-let-7b-5p, hsa-miR-27a-3p, hsa-let-7c-5p, and hsa-miR-182-5p. The final model, integrating clinical data and miRNA predictors, achieved an adjusted LOOCV-AUC of 0.940 (95% CI 0.861-1.000), outperforming the model with only clinical data (AUC 0.719, 95% CI 0.557-0.868). Conclusion: A combined model based on miRNA and clinical data showed promising results for early assessment of bronchopulmonary dysplasia risk, achieving a cross-validated AUC of 0.940 in a small cohort. A correlation-based sensitivity analysis reduced systemic impact, enabling the identification of specific candidate markers. These results require external validation and confirmation by RT-qPCR before clinical implementation.