Predicting Parkinson disease related genes based on PyFeat and gradient boosted decision tree

Area of Science:

• Biomedical analysis
• Genetics
• Bioinformatics

Background:

• Identifying genes related to Parkinson's disease (PD) is crucial for diagnosis and treatment.
• Existing methods often overlook long noncoding RNA (lncRNA) genes, which are important in disease development.
• There is a need for systems that can predict both protein and lncRNA genes associated with PD.

Purpose of the Study:

• To propose a novel prediction system for identifying protein and lncRNA genes related to Parkinson's disease (PD).
• To aid in the early diagnosis of PD by identifying disease-associated genes.
• To evaluate the system's performance using multiple metrics and compare it with existing techniques.

Main Methods:

• Genes were preprocessed into DNA FASTA sequences from the UCSC genome browser.
• Significant DNA sequence features were extracted using the PyFeat method, with AdaBoost for feature selection.
• A gradient-boosted decision tree (GBDT) model was employed for gene prediction and diagnosis.

Main Results:

• The proposed system achieved promising results, with an average accuracy of 78.6%.
• Key performance metrics included an area under the curve (AUC) of 84.5%, AUPR of 85.3%, and F1-score of 78.3%.
• Predicted top-rank protein and lncRNA genes were validated through a literature review, confirming their relevance to PD.

Conclusions:

• The developed system effectively predicts both protein and lncRNA genes associated with Parkinson's disease.
• The approach demonstrates superior performance compared to other existing systems.
• This work contributes to early PD diagnosis and provides valuable insights for future research into PD-related genes.