Dna methylation quality control

dna-methylation-quality-control

Summary

Apply detection p-value and bead count thresholds to remove low-quality probes from Illumina methylation array data (450K or EPIC). This filtering step is essential for downstream analysis, removing probes with insufficient signal reliability before normalization and differential methylation analysis.

When to use

Apply this skill immediately after loading raw .idat files or beta-value matrices from HumanMethylation450 or EPIC arrays when you need to exclude probes that fail quality control. Specifically, use it when your input dataset contains detection p-values and bead count information and you have not yet performed downstream analyses (normalization, batch correction, or DMR detection).

When NOT to use

• Input data has already been filtered by another pipeline or tool (detection p-values and bead counts no longer available)
• You are working with single-cell methylation data or non-array-based methods (WGBS, bisulfite sequencing)
• Your analysis explicitly requires retaining low-signal probes for specific methodological reasons

Inputs

• Raw .idat files from Illumina methylation array (450K or EPIC)
• Beta-value matrix with detection p-values and bead count data
• Sample metadata (phenotype information, batch labels)

Outputs

• Filtered probe count matrix (probes × samples)
• Quality control report documenting probe removal statistics
• Pre- and post-filter probe count comparison
• Bead count distribution plots

How to apply

Load the methylation array data using ChAMP data import functions (from .idat files or beta-valued matrix), then apply champ.filter() with default parameters. This function performs two successive filtering steps: (1) removal of probes with detection p-value > 0.01, and (2) removal of probes with fewer than 3 beads in at least 5% of samples per probe. Compare pre- and post-filter probe counts and examine bead count distributions to verify filtering efficacy. Document the number of probes retained and removed in a quality control report.

Related tools

• ChAMP (Primary tool for filtering probes via champ.filter() function; provides comprehensive pipeline from data loading through quality control to differential methylation analysis) — https://github.com/YuanTian1991/ChAMP
• ChAMPdata (Dependency package providing methylation array annotations and test datasets (HumanMethylation450, EPIC) required to run ChAMP filtering) — https://github.com/YuanTian1991/ChAMPdata
• minfi (Alternative package for 450K and EPIC array analysis; offers Functional Normalization and data import methods)
• RnBeads (Alternative pipeline for 450K or EPIC array analysis with comparable quality control and filtering capabilities)

Examples

library(ChAMP); champ.filter(beta = your_beta_matrix, pd = your_sample_metadata)

Evaluation signals

• Pre-filter probe count > post-filter probe count (probes successfully removed)
• Number of removed probes with detection p-value > 0.01 matches or exceeds expected frequency in the dataset
• Number of probes removed due to bead count < 3 in ≥5% of samples is documented and reasonable given sample size
• Post-filter probe matrix contains no probes with detection p-value > 0.01 (spot-check random probes)
• Bead count distribution before filtering shows probes below threshold; post-filter distribution shows threshold is respected

Limitations

• Filtering thresholds (detection p-value > 0.01, bead count < 3 in ≥5% samples) are defaults that may not be optimal for all tissue types or array batches; users may need to adjust based on data characteristics
• Filtering requires that raw detection p-values and bead count data are preserved in the input object; some upstream pipelines may have already discarded this information
• The HumanMethylation450 test dataset contains only 8 samples (4 tumor, 4 control), which may not reflect filtering behavior in larger, more diverse cohorts
• ChAMP development version on GitHub is 'under intensive modification and upgrade'; formally released stable versions are maintained on Bioconductor

Evidence

• [other] champ.filter() applies two successive filtering steps by default: (1) removal of probes with detection p-value > 0.01, and (2) removal of probes with fewer than 3 beads in at least 5% of samples per probe.: "champ.filter() applies two successive filtering steps by default: (1) removal of probes with detection p-value > 0.01, and (2) removal of probes with fewer than 3 beads in at least 5% of samples per"
• [readme] ChAMP package is designed for conduct DNA methylation array analysis, providing service from data loading, to final gene set enrichment analysis: "ChAMP package is designed for conduct DNA methylation array analysis, providing service from data loading, to final gene set enrichment analysis"
• [intro] a variety of different data import methods (e.g. from .idat files or a beta-valued matrix): "a variety of different data import methods (e.g. from .idat files or a beta-valued matrix)"
• [intro] The 450k lung tumor data set contains only 8 samples, 4 lung tumor samples (T) and 4 control samples (C): "The 450k lung tumor data set contains only 8 samples, 4 lung tumor samples (T) and 4 control samples (C)"
• [readme] Note that this is NOT a proper release version ChAMP and under intensive modification and upgrade, the formally released one is on Bioconductor: "Note that this is NOT a proper release version ChAMP and under intensive modification and upgrade, the formally released one is on Bioconductor"