illumina-array-data-processing-epic

Illumina array data processing (EPIC)

Summary

End-to-end processing pipeline for Illumina EPIC methylation array data, from raw .idat files through quality control, normalization, batch correction, and differentially methylated region detection. This skill encompasses the complete workflow for analyzing 450k and EPIC array methylation data using ChAMP, a comprehensive R package integrating multiple normalization and statistical methods.

When to use

You have raw Illumina EPIC or 450k methylation array data (.idat files or beta-valued matrices) and need to perform comprehensive quality assessment, probe correction, batch effect adjustment, and identification of differentially methylated regions or blocks across sample groups.

When NOT to use

• Input data are already normalized and batch-corrected; use only downstream differential methylation detection methods instead.
• Array type is not EPIC, 450k, EPICv2, or mouse arrays; ChAMPdata does not support other platforms.
• Sample size is extremely small (e.g., n=1–2 per group); statistical power for DMR detection will be insufficient.

Inputs

• .idat files (raw Illumina array intensity data)
• Beta-valued methylation matrix
• Sample phenotype/group metadata
• ChAMPdata annotation objects (e.g., AnnoEPIC, AnnoEPICv2)

Outputs

• Quality control plots and metrics
• Normalized beta-value matrix
• Batch-corrected methylation data
• Differentially methylated region (DMR) calls with statistical annotations
• Differentially methylated block detection results
• Copy-number alteration inference results

How to apply

Load your .idat files or beta-valued matrix using ChAMP's data import functions (e.g., champ.load()). Apply Type-2 probe correction using SWAN, PBC, or BMIQ (BMIQ is default). Perform Functional Normalization from minfi or implement SVD-based batch effect analysis. Correct for multiple batch effects using ComBat if needed. Adjust for cell-type heterogeneity via RefbaseEWAS when applicable. For differential methylation detection, choose between Probe Lasso, Bumphunter, or DMRcate for DMR identification, or use champ.Block() for differentially methylated block detection. Verify no spurious blocks are detected by examining output with Block.GUI() interactive interface when using simulation data (e.g., EPICSimData) to confirm expected behavior.