Data processing choices can affect findings in differential methylation analyses: an investigation using data from...

Area of Science:

• Epigenetics
• Bioinformatics
• Genomics

Background:

• Numerous methods exist for processing and analyzing DNA methylation (DNAm) data.
• Standardized comparisons are crucial for understanding method behavior and reliability.

Purpose of the Study:

• To systematically compare DNA methylation data processing and analysis methods.
• To evaluate method performance in detecting true effects and spurious findings.
• Utilized cord blood DNAm data from the LIMIT Randomized Controlled Trial (RCT).

Main Methods:

• Analyzed DNAm data from 645 cord blood samples using Illumina 450K BeadChip arrays.
• Compared three normalization methods, pre- or post-probe filtering.
• Evaluated supervised, unsupervised, and model-adjusted batch effect handling.
• Assessed analysis with and without adjustment for estimated cell type proportions.

Main Results:

• Significant differences observed across various data processing and analysis choices.
• Certain processing combinations led to a substantial number of spurious 'significant' findings.
• Method choices influenced the number and identity of differentially methylated probes.

Conclusions:

• Data processing and analysis choices critically affect DNA methylation study outcomes.
• Emphasize the importance of sensitivity analyses and replication in epigenetics research.
• Recommendations for robust DNA methylation data analysis and interpretation.