Identification of predictive miRNA biomarkers from neonatal CSF extracellular vesicles in post-hemorrhagic ventricular dilatation via machine learning approaches

Introduction:
Neonatal post-hemorrhagic ventricular dilatation (PHVD), a serious complication of intraventricular hemorrhage (IVH), is linked to significant neurodevelopmental impairments. Early biomarkers are critical for timely prognosis and intervention. At the onset of PHVD cerebrospinal fluid (CSF) is drained to reduce the intercranial pressure, we aim to use this abundantly available biofluid to identify miRNA biomarkers from isolated extracellular vesicles (EVs).

Methods:
We isolated EVs from 190 CSF samples from 43 neonates diagnosed with PHVD. High-throughput sequencing was used to profile EV-associated miRNAs, generating a novel dataset of miRNA signatures. Machine learning (ML) and artificial intelligence (AI) algorithms were applied to analyze these miRNA profiles and predict neurodevelopmental outcomes at 2 and 3 years of age.

Results:
We were able to detect 248 miRNAs consistently in 80% of the samples. Our ML models identified key miRNAs with strong predictive power for cognitive and motor outcomes. These miRNAs showed potential as early, minimally invasive biomarkers for neurodevelopmental prognosis. The identified biomarkers provided insights into the pathophysiology of PHVD and may inform the development of targeted therapies.

Summary/Conclusions:
This study highlights the potential of EV-associated miRNAs as predictive biomarkers for neurodevelopmental outcomes in neonates with PHVD. Further validation in larger cohorts could establish these miRNAs as valuable clinical tools for early diagnosis and prognosis, offering new avenues for mitigating long-term neurodevelopmental deficits.