| dc.description.abstract | Background: Huntington disease (HD) is a rare neurodegenerative fatal disorder. HD is caused by an autosomal dominant mutation in the Huntingtin protein causing extension of CAG repeats. Although this mutation was identified in 1993, no disease-modifying treatments are currently available; existing therapies only manage symptoms. To optimize treatment effectiveness before irreversible brain damage occurs, it is crucial to better understand early disease pathogenesis and to identify outcome measures that are sensitive to changes during the presymptomatic phase. Currently, the early benefit of therapeutic interventions cannot be reliably assessed due to the lack of sensitive biomarkers capable of detecting subtle presymptomatic changes.
Aim: This study aimed to investigate whole blood mRNA in presymptomatic and early symptomatic HD patients, with the hypothesis that there are different stages prior symptoms’ onset. The goal is to identify biomarkers capable of differentiating between presymptomatic disease states.
Methods: A machine learning method, recursive ensemble feature selection (REFS) in a nested tenfold cross validation scheme was applied to a transcriptomic dataset from the TRACK-HD study. This methodology was chosen to address challenges related to overfitting and bias commonly associated with high-dimensional datasets and small sample sizes. The dataset used in the study comprises 54,675 features and were classified into three groups: early presymptomatic (preHD-A, n=17), late presymptomatic (preHD-B, n=18), and early symptomatic (Stage 1, n=18). Feature analysis was conducted using RStudio with Bioconductor packages and web-based tools.
Results: The REFS model achieved high classification accuracy with low features among the three patients’ groups. Specifically, preHD-A and preHDB were distinguished with 96.7% accuracy using 42 features. Classification accuracy remained high (93.9%) even when the feature set was reduced to four transcripts: LAMA4, ZIC5, smAKAP, and one unlabeled transcript.
Conclusion: This study demonstrates that distinct transcriptomic profiles exist within the presymptomatic phase of HD. We propose four gene expression features as potential biomarkers for disease progression toward the symptomatic stage. Further investigation into these markers may contribute to the development of early therapeutic interventions and/or pave the way for assessment of treatment efficacy in presymptomatic HD. | |
