Disease classification for whole-blood DNA methylation: Meta-analysis, missing values imputation, and XAI.

Area of Science:

• Genomics and Bioinformatics
• Computational Biology
• Epigenetics

Background:

• DNA methylation plays a crucial role in gene expression and is linked to various diseases.
• Meta-analysis of diverse DNA methylation datasets necessitates a specialized workflow for integrated data processing.

Approach:

• Developed a comprehensive computational approach for classifying controls and patients using combined DNA methylation datasets.
• Integrated data harmonization, machine learning model construction, dimensionality reduction, missing value imputation, and explainable AI (XAI) for model interpretation.

Key Points:

• Data harmonization improved classification accuracy by up to 20% when training and test datasets used different preprocessing methods.
• Tree ensemble models achieved over 95% accuracy for Parkinson's disease classification.
• Dimensionality reduction and imputation methods maintained high classification accuracy while simplifying models and handling missing data.

Conclusions:

• The proposed method offers a robust approach for classifying healthy individuals and patients using whole-blood DNA methylation data, exemplified by Parkinson's disease and schizophrenia.
• The algorithm effectively addresses data harmonization challenges for large-scale meta-analyses, imputes missing values, and constructs low-dimensional classification models.