A framework for analyzing DNA methylation data from Illumina Infinium HumanMethylation450 BeadChip

Area of Science:

• Epigenetics and Genomics
• Computational Biology
• Biostatistics

Background:

• DNA methylation is strongly linked to complex human diseases.
• The Illumina Infinium HumanMethylation450 BeadChip (450K) is a key technology for profiling DNA methylation.
• Analysis of 450K array data faces challenges due to inherent biases.

Purpose of the Study:

• To establish a comprehensive framework for evaluating DNA methylation data analysis methods.
• To guide researchers in selecting optimal analytical approaches for Illumina 450K array data.
• To address challenges in DNA methylation data analysis, particularly biases.

Main Methods:

• Developed a generalized framework encompassing data import, quality control, normalization, bias correction, differential methylation detection, and biological interpretation.
• Evaluated five distinct analysis methods using three real-world human DNA methylation datasets.
• Assessed the performance of methods across various stages of the analysis pipeline.

Main Results:

• Identified Minfi and methylumi as optimal choices for analyzing smaller DNA methylation datasets.
• Determined BMIQ and RCP as effective methods for correcting type bias in 450K data.
• Confirmed that missMethyl is suitable for Gene Ontology (GO) term enrichment and biological interpretation of differentially methylated probes (DMPs).

Conclusions:

• The proposed framework provides a systematic approach to DNA methylation data analysis.
• Specific R packages (Minfi, methylumi, BMIQ, RCP, missMethyl) are recommended for different analytical tasks.
• These recommendations facilitate more robust and reliable identification of disease-associated epigenetic variations.