BayMeth: Improved DNA methylation quantification for affinity capture sequencing data using a flexible Bayesian approach
DNA methylation (DNAme) is a critical component of the epigenetic regulatory machinery and aberrations in DNAme patterns occur in many diseases, such as cancer. Mapping and understanding DNAme profiles offers considerable promise for reversing the aberrant states. There are several approaches to analyze DNAme, which vary widely in cost, resolution and coverage. Affinity capture and high-throughput sequencing of methylated DNA strike a good balance between the high cost of whole genome bisulphite sequencing (WGBS) and the low coverage of methylation arrays. However, existing methods cannot adequately differentiate between hypomethylation patterns and low capture efficiency, and do not offer flexibility to integrate copy number variation (CNV). Furthermore, no uncertainty estimates are provided, which may prove useful for combining data from multiple protocols or propagating into downstream analysis. We propose an empirical Bayes framework that uses a fully methylated (i.e. SssI treated) control sample to transform observed read densities into regional methylation estimates. In our model, inefficient capture can be distinguished from low methylation levels by means of larger posterior variances. Furthermore, we can integrate CNV by introducing a multiplicative offset into our Poisson model framework. Notably, our model offers analytic expressions for the mean and variance of the methylation level and thus is fast to compute. Our algorithm outperforms existing approaches in terms of bias, mean-squared error and coverage probabilities as illustrated on multiple reference datasets. Although our method provides advantages even without the SssI-control, considerable improvement is achieved by its incorporation. Our method can be applied to methylated DNA affinity enrichment assays (e.g MBD-seq, MeDIP-seq) and a software implementation is available in the Bioconductor Repitools package.