Comparison of pre-processing methodologies for Illumina 450k methylation array data in familial analyses
RESULTS: Stratified quantile normalisation combined with ComBat were consistently found to be the most appropriate when assessed using density, MDS and cluster plots. This was supported quantitatively by ANOVA on the first principal component where the effect of batch dropped from p < 0.01 to p = 0.97 after stratified QN and ComBat. Median absolute differences between replicated samples were the lowest after stratified QN and ComBat as were the standard error measures on known imprinted regions. Biological information was preserved after normalisation as indicated by the maintenance of a significant association between a known mQTL and methylation (p = 1.05e-05).
CONCLUSIONS: A strategy combining stratified QN with ComBat is appropriate for use in the analyses when no reference sample is available but preservation of biological variation is paramount. There is great potential for use of 450k array data to further our understanding of the methylome in a variety of similar settings. Such advances will be reliant on the determination of appropriate methodologies for processing these data such as established here.
Publication titleClinical Epigenetics
Department/SchoolMenzies Institute for Medical Research
PublisherBioMed Central Ltd.
Place of publicationUnited Kingdom
Rights statementCopyright 2016 The Authors. Licensed under Creative Commons Attribution 4.0 International (CC BY 4.0) https://creativecommons.org/licenses/by/4.0/