Defining, Validating, and Endotyping Phenotypes in BPD - Bronchopulmonary dysplasia (BPD) is a complex and heterogeneous lung disease affecting preterm infants, leading to long-term respiratory and neurodevelopmental complications. Current classification methods rely on clinical observations at 36 weeks postmenstrual age, limiting early identification of high-risk infants and delaying targeted interventions. The absence of biologically defined BPD subtypes further hinders the development of personalized treatment strategies, restricting our ability to tailor care. This project seeks to address these critical gaps by integrating artificial intelligence-driven phenotyping with transcriptomic profiling to define and validate clinically and biologically meaningful BPD subgroups. On the basis of our novel preliminary findings, we hypothesize that BPD consists of distinct clinical phenotypes that can be identified through machine learning-based clustering and validated across external datasets, and that these phenotypes are driven by unique molecular endotypes. Furthermore, we hypothesize that transcriptomic profiling will reveal key biological pathways underlying phenotype-specific disease progression, providing mechanistic insights into BPD heterogeneity and potential molecular targets for intervention. To test this hypothesis, we will analyze a large NICHD neonatal database of 3,198 at-risk infants and external contemporary datasets using advanced computational methods. Aim 1 will establish clinically distinct BPD phenotypes using machine learning-based computational modeling techniques, incorporating perinatal history, respiratory support parameters, comorbidities, and early disease trajectories. Aim 2 will validate these phenotypes using external datasets reflecting modern neonatal intensive care unit practices and assess their alignment with historical classifications. Aim 3 will define transcriptomic endotypes associated with validated BPD phenotypes by leveraging RNA sequencing of neonatal blood samples, revealing key biological pathways underlying disease progression and identifying molecular targets for precision medicine approaches. By integrating large-scale clinical data, artificial intelligence, and molecular profiling, this study will provide a comprehensive framework for understanding BPD heterogeneity at both clinical and biological levels. The findings have the potential to transform our understanding of BPD pathogenesis, improve early risk prediction, and lay the foundation for phenotype-specific interventions in preterm infants. The collaborative team, composed of experts in neonatology, computational phenotyping, epidemiology and biostatistics, and bioinformatics, is uniquely positioned to execute this study with rigor, ensuring a transformative impact on neonatal health.
