Tutorial for EPIC v1 and EPIC v2 joint methylation data analysis in RnBeads

Introduction

This vignette demonstrates a cross-platform analysis of DNA methylation data that involves methylation calls from Illumina EPIC v1 and EPIC v2 microarrays.

Example data set

This tutorial utilizes data released from an EPIC v2 validation study (Kaur et al.). The IDAT files are publicly available at the Gene Expression Omnibus (GEO) database. Technical replicate samples from cell lines listed below were provided with EPIC v1 and EPIC v2 data.

• HCT116

  • Replicate 1 (GSM7139680)

  • Replicate 2 (GSM7139681)

• GM12878

  • Replicate 1 (GSM7139664)

  • Replicate 2 (GSM7139667)

Load RnBSet objects

Provided RnBSet objects of EPIC v1 and EPIC v2 data were created by using the updated RnBeads.hg38 annotation package.

EPIC v1

	HCT116	GM12878
Replicate 1	HCT116_Rep1_EPICv1	GM12878_Rep1_EPICv1
Replicate 2	HCT116_Rep2_EPICv1	GM12878_Rep2_EPICv1

EPIC v2

	HCT116	GM12878
Replicate 1	HCT116_Rep1_EPICv2	GM12878_Rep1_EPICv2
Replicate 2	HCT116_Rep2_EPICv2	GM12878_Rep2_EPICv2

Load the RnBSet objects by using the following code.

library(RnBeads)
library(RnBeads.hg38)

rnbset_path <- "path/to/rnbset"

## EPIC v1
HCT116_Rep1_EPICv1 <- load.rnb.set(file.path(analysis_path, "HCT116_Rep1_EPICv1.RDS"))
HCT116_Rep2_EPICv1 <- load.rnb.set(file.path(analysis_path, "HCT116_Rep2_EPICv1.RDS"))

GM12878_Rep1_EPICv1 <- load.rnb.set(file.path(analysis_path, "GM12878_Rep1_EPICv1.RDS"))
GM12878_Rep2_EPICv1 <- load.rnb.set(file.path(analysis_path, "GM12878_Rep2_EPICv1.RDS"))

## EPIC v2
HCT116_Rep1_EPICv2 <- load.rnb.set(file.path(analysis_path, "HCT116_Rep1_EPICv2.RDS"))
HCT116_Rep2_EPICv2 <- load.rnb.set(file.path(analysis_path, "HCT116_Rep2_EPICv2.RDS"))

GM12878_Rep1_EPICv2 <- load.rnb.set(file.path(analysis_path, "GM12878_Rep1_EPICv2.RDS"))
GM12878_Rep2_EPICv2 <- load.rnb.set(file.path(analysis_path, "GM12878_Rep2_EPICv2.RDS"))

Combining the RnBSet objects

Method combine() has now been adapted to allow the merger of any two RnBeads data container objects. It has an important parameter type that specifies the set operation for merging the methylation sites:

• type="all" will include all sites present in both objects

• type="all.x" will retain all sites present in the first object

• type="all.y" will retain all sites present in the second object

• type="common" will retain only sites found in both objects

Let us combine all data sets and perform a sequence of joint cross-platform analyses of all data. First combine the two array-base data sets:

## Combine HCT116 Replicate 1
HCT116_Rep1_combined <- rnb.combine.arrays(HCT116_Rep1_EPICv1, HCT116_Rep1_EPICv2, type = "common")

This creates an RnBeadRawSet object with 2 samples that contain 721435 common probes. Probe type counts are: 718599 CpG probes, 2779 CpH probes, 57 rs probes and 0 nv probes.

The other replicates can be combined the same way as the code above shows.

Joint analysis of the combined data

Make sure the RnBeads assembly option is set to hg38. This ensures the usage of the RnBeads.hg38 annotation package. Additionally, for differential analysis the following settings shall be configured as below. Please note that min.group.size option is set to 1 for this tutorial since the analysis will be conducted for each replicate sample.

rnb.options(assembly = "hg38")

## Differential comparison options
rnb.options(differential.comparison.columns = c("EPIC_version"))
rnb.options(columns.pairing = NULL)
rnb.options(min.group.size = 1)
rnb.options(differential.report.sites = TRUE)
rnb.options(identifiers.column="SampleID")

The RnBeads analysis pipeline can be run simply by using this new combined data set.

rnb.options(analysis.name = "EPICv1_vs_EPICv2_HCT116_Replicate_1")
report.dir <- file.path("path/to/HCT116_Rep1_Report")

res = rnb.run.analysis(data.source = HCT116_Rep1_combined, dir.report=report.dir)

The differential analysis scatterplot shows a high correlation number between the samples of two platforms (Figure 1.).

Figure 1: Scatterplot for differential methylation of EPIC v1 and EPIC v2 probes - HCT116 / Replicate 1

Kaur et al. have demonstrated that some of these differentially methylated probes show altered Infinium probe designs and strands.

Overall, it’s difficult to argue that the Infinium design changes and the different strand types of the common probes are enough to influence the differential methylation characteristics of these probes.

Conclusion

The recent RnBeads update enables a combined analysis of DNA methylation data from EPIC v1 and EPIC v2 platforms using the RnBeads.hg38 annotation package. The results show the consistency of DNA methylation levels that are measured by the common probes of EPIC v1 and EPIC v2 platforms.

References

Kaur, D., Lee, S. M., Goldberg, D. J., Spix, N. J., Hinoue, T., Li, H., Dwaraka, V. B., Smith, R., Shen, H., Liang, G., Renke, N., Laird, P. W., & Zhou, W. (2023). Comprehensive evaluation of the Infinium human MethylationEPIC v2 BeadChip. Epigenetics Communications, 3(1). https://doi.org/10.1186/s43682-023-00021-5