Coverage recommendations for methylation analysis by whole-genome bisulfite sequencing

Abstract

Whole-genome bisulfite sequencing (WGBS) allows genome-wide DNA methylation profiling, but the associated high sequencing costs continue to limit its widespread application. We used several high-coverage reference data sets to experimentally determine minimal sequencing requirements. We present data-derived recommendations for minimum sequencing depth for WGBS libraries, highlight what is gained with increasing coverage and discuss the trade-off between sequencing depth and number of assayed replicates.

Figure 1. Coverage requirements for WGBS experiments

1. Heatmap showing the pairwise Pearson correlation coefficients (PCC) for genome-wide methylation profiles of the samples used in this study (n=14). Average methylation levels were estimated in 1kb tiling windows.

2. Distribution of DMR sizes (x-axis) and average methylation difference (y-axis) for DMRs found at 30× comparing hESCs to human cortex (red), CD184 to liver (grey) and CD4 to CD8 (blue) using 2 replicates each. Black dot indicates median and ellipsoids span from the 25^th to the 75^th percentile in each dimension.

3. True positive rate (TPR, y-axis) as a function of coverage (x-axis) for hESC vs. cortex (red), CD184 vs. liver (grey) and CD4 vs. CD8 (blue) using 2 replicates for each group. True positive rate is defined as the fraction of high coverage (30×) reference DMRs recovered at the coverage level indicated. Grey box indicates coverage range where change in TPR exhibits the largest drop.

4. Distribution of DMR sizes (x-axis) and average methylation difference (y-axis) for DMRs discovered at 1× (grey) and additional DMRs discovered when increasing the coverage from 1× to 5× (dark red), 5× to 10× (light red) and 10× to 30× (orange) in the hESCs vs. human cortex comparison using 2 replicates each. Black dots indicate median and ellipsoids span from the 25^th to the 75^th percentile in each dimension.

5. False discovery rate (FDR, y-axis) as function of coverage (x-axis) for DMRs exhibiting a methylation difference of 20% or greater when comparing hESCs to human cortex (red), CD184 to liver (grey) or CD4 to CD8 (blue) using two replicates for each group.

Figure 2. Replicate recommendations

1. True positive rate (TPR, y-axis) as a function of coverage (x-axis) comparing hESCs to human cortex using 1, 2 or 3 replicates per group for DMRs with a methylation difference greater than 20%.

2. False discovery rate (FDR, y-axis) as a function of coverage (x-axis) comparing hESCs to human cortex using 1, 2 or 3 replicates per group for DMRs with a methylation difference greater than 20%.

3. Percentage of 3 replicate based, reference DMRs with a methylation difference greater than 20% that are recovered as a function of total coverage used for the entire experiment. Lines indicate whether total experimental coverage is distributed across 1, 2 or 3 replicates