Predictive modelling of Parkinson's disease progression based on RNA-Sequence with densely connected deep recurrent...

Area of Science:

• Genomics
• Neuroscience
• Computational Biology

Background:

• High-throughput sequencing generates vast genomic and transcriptomic data relevant to Parkinson's disease (PD).
• Existing predictive models for PD progression primarily utilize longitudinal clinical data, with a gap in RNA-sequence data-based approaches.
• Parkinson's Progression Markers Initiative (PPMI) provides valuable longitudinal RNA-Seq data for PD research.

Purpose of the Study:

• To develop a predictive model for forecasting Parkinson's disease progression at a patient's next medical visit.
• To leverage longitudinal RNA-Sequence data and capture temporal patterns for PD progression prediction.
• To establish a novel computational approach for PD progression assessment using transcriptomic data.

Main Methods:

• Utilized RNA-Sequence data from 423 Parkinson's disease patients from the PPMI dataset.
• Developed a deep Recurrent Neural Network (RNN) architecture incorporating dense connections and batch normalization.
• Analyzed longitudinal data spanning 4 years with 34,682 predictor variables.

Main Results:

• The proposed deep RNN model accurately predicts PD progression using high-dimensional RNA-Seq data.
• Achieved a Root Mean Square Error (RMSE) of 6.0 for progression prediction.
• Demonstrated a significant rank-order correlation (r=0.83, p<0.0001) between predicted and actual PD disease status.

Conclusions:

• Deep learning models can effectively predict Parkinson's disease progression from RNA-Seq data.
• This approach offers a novel method for monitoring and potentially managing PD.
• RNA-Seq data holds significant potential for advancing our understanding and prediction of neurodegenerative disease progression.