Objective: To investigate the role of histone modifications in Juvenile Huntington’s Disease (JHD) using machine learning, aiming to identify novel epigenetic biomarkers predictive of disease onset and severity.

Background: However, the profound clinical variability in disease onset and progression strongly implicates secondary epigenetic regulatory mechanisms. Understanding specific histone modifications, which dynamically govern chromatin accessibility, may provide critical mechanistic insights into JHD pathogenesis and unveil new disease-modifying intervention strategies.

Method: This study analyzed Chromatin Immunoprecipitation Sequencing (ChIP-seq) data from neuronal cells of 250 JHD patients and age-matched controls. We targeted histone marks regulating gene expression, including H3K4me3 and H3K27ac. Whole-genome sequencing was integrated with these epigenetic datasets to correlate histone landscapes with clinical parameters. A Convolutional Neural Network (CNN) was deployed to extract spatial patterns within histone modification peaks. Subsequently, a Random Forest (RF) algorithm integrated these features with genomic data for predictive modeling. Model generalizability was rigorously validated using stratified cross-validation.

Results: JHD patients exhibited distinct histone modification profiles correlating with accelerated clinical decline. Specifically, increased H3K27ac levels were enriched near the HTT gene and loci driving neuroinflammation and synaptic dysfunction. Our hybrid CNN-RF pipeline achieved 88% predictive accuracy in stratifying early- versus late-onset JHD cohorts, demonstrating an Area Under the Curve (AUC) of 0.94. The model accurately predicted the exact age of disease onset with a Mean Absolute Error of 2.3 years. Crucially, these specific epigenetic alterations were mathematically associated with a 30% acceleration in overall disease progression.

Conclusion: Machine learning-driven profiling of histone modifications elucidates the complex epigenetic landscape of JHD. The identification of predictive markers like H3K27ac provides a highly accurate algorithmic approach to forecasting disease onset, paving the way for targeted, personalized therapeutic interventions in neurodegeneration.

« Back to 2026 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/machine-learning-driven-analysis-of-epigenetic-signatures-in-juvenile-huntingtons-disease-investigating-histone-modification-dynamics/