A multilevel model to address batch effects in copy number estimation using SNP arrays.

Area of Science:

• Genetics and Genomics
• Bioinformatics
• Statistical Genetics

Background:

• Submicroscopic changes in chromosomal DNA copy number are linked to diseases and cancers.
• High-throughput technologies detect segmental DNA copy number changes, but are susceptible to batch effects.
• Batch effects can lead to incorrect inferences in genomewide array analyses.

Purpose of the Study:

• To develop a robust method for estimating DNA copy number variations (CNVs).
• To explicitly model and correct for batch effects in genomewide array data.
• To improve locus-specific estimates of copy number uncertainty using shrinkage.

Main Methods:

• Extended previous model-based approaches for copy number estimation.
• Explicitly modeled batch effects and used shrinkage for improved estimates.
• Utilized biallelic genotype calls to estimate batch- and locus-specific parameters without training data.

Main Results:

• Demonstrated robustness in studies of bipolar disease and chromosome 21 trisomy.
• Successfully adjusted for dominant batch effects in quantile-normalized intensities.
• Showcased accuracy with datasets containing significant copy number alterations.

Conclusions:

• The developed approach effectively models batch effects for more accurate CNV detection.
• Locus-specific estimates aid in investigating mosaicism and guiding downstream analyses.
• Open-source software (R package crlmm) is available for broader application.