Meffil

Meffil performs scalable quality control, memory-efficient functional normalization, and preparation for epigenome-wide association studies (EWAS) on large-scale DNA methylation datasets generated by Illumina Methylation BeadChip microarrays, including Infinium HumanMethylation450 and MethylationEPIC BeadChips.

Key Features:

• Dataset support: Processes DNA methylation data from Illumina Methylation BeadChip microarrays, including Infinium HumanMethylation450 and MethylationEPIC BeadChips.
• Scalability: Handles large datasets in sample size and genome coverage to accommodate extensive methylation studies.
• Memory-efficient functional normalization: Implements a complete re-implementation of functional normalization that minimizes computational memory usage while maintaining processing speed.
• Mixed-effects normalization: Integrates fixed and random effects within the functional normalization procedure.
• Automated parameter estimation: Automatically estimates normalization parameters to reduce technical variation in DNA methylation levels.
• Distributed normalization: Supports normalizing datasets distributed across different physical locations without sharing biologically-based individual-level data.
• QC and EWAS preparation: Provides functionality for quality control and preparation of data for epigenome-wide association studies.

Scientific Applications:

• Quality control of methylation arrays: Performing QC on Illumina BeadChip methylation data prior to downstream analysis.
• Normalization of Infinium arrays: Applying functional normalization, including fixed and random effects, to Infinium HumanMethylation450 and MethylationEPIC data.
• Epigenome-wide association studies (EWAS): Preparing large-scale methylation datasets for EWAS to reduce false positives and increase analytical power.
• Meta-analysis harmonization: Reducing heterogeneity in meta-analyses by enabling distributed normalization without sharing individual-level biological data.

Methodology:

Computational methods include a complete re-implementation of functional normalization with integrated fixed and random effects, automated estimation of normalization parameters, memory-usage minimization, and support for distributed normalization across physical locations without sharing individual-level biological data.