Crossreactive probes on Illumina DNA methylation arrays: a large study on ALS shows that a cautionary approach is warranted in interpreting epigenome-wide association studies

Scripts accompanying the manuscript 'Crossreactive probes on Illumina DNA methylation arrays: a large study on ALS shows that a cautionary approach is warranted in interpreting epigenome-wide association studies'. This repository contains the scripts used for analyses. For convenience, we wrapped several functions into an R package, which is available at: https://github.com/pjhop/DNAmCrosshyb.

Data availability

The data used in downstream analyses can be downloaded from: https://doi.org/10.5281/zenodo.4110015. All downstream analyses in c9_analysis.Rmd (https://github.com/pjhop/dnamarray_crossreactivity/blob/master/analysis/c9_analysis.Rmd) and in supplementary_note.Rmd (https://github.com/pjhop/dnamarray_crossreactivity/blob/master/analysis/supplementary_note.Rmd) can be reproduced using the deposited data as follows:

• Clone this repository: < git clone https://github.com/pjhop/dnamarray_crossreactivity.git >
• Download the data ('data.zip') and place it in the 'dnamarray_crossreactivity' folder.
• Unzip the data.zip folder

Overview

R: Contains some reusable scripts for plotting and mapping probes to sequences.
analysis: Contains scripts used for the analyses presented in the paper

• preprocessing: Folder that contains several scripts used for preprocessing the data. This includes transforming the data to the format used in the OSCA software, PCA, extracting signal intensities and normalizing out-of-band (OOB) signal intensities.
• ewas: Folder that contains scripts to perform EWAS analyses, most of which is done using the OSCA software.
  Also contains scripts to perform meta-analyses.
• c9_matches: Folder that contains scripts to map 450k and EPIC probes to the C9orf72 repeat expansion. Includes scripts to do so in different ways, for example: assuming that the repeat is methylated/unmethylated, representing CpGs by YpG, including flanking regions etc.
• other: Folder that contains script to match 450k/EPIC probes to various repeat sequences.
• c9_analysis.Rmd: Main Rmarkdown file in which all downstream analyses are performed and figures/tables are generated.
• supplementary_note.Rmd: Rmarkdown that generates the supplementary note.

R

ewas_plot_functions.R: Contains several plotting functions (e.g. qqplots, manhattan plots, comparing tstats). See the markdown-file for usage.
probe_mapper_nonref.R: Map probes to a non-reference sequence. This function is used in the scripts in the analysis/c9_matches folder. Note that an updated version is included as a function in the DNAmCrosshyb package, called map_probes_sequence.
create_binary_files.R: R-wrapper to create binary files using the OSCA software.

analysis/c9_analysis.Rmd

analysis/preprocessing

create_binary_files_*: Convert matrices containing beta-values/m-values/signal intensities to binary format used in the OSCA software.
pca_*: Perform PCA using OSCA software.
naten_*: Extract and normalize out-of-band (OOB) signal intensities. The get_OOB() function from the DNAmCrosshyb package is used for this.
save_total_intensities_NL.R: Extract total signal intensities and normalize.

analysis/ewas

mlma_loco_*: Perform a mixed linear model (leave-one-chromosome out) as implemented in the OSCA software. Different sets of samples/covariates are used for sensitivity analyes (e.g. removing PCA outliers, using m-values instead of beta-values etc.).
mlma_linear_*: Performed a linear model using the OSCA software.
save_covariates_*: Save sets of covariates used in the different OSCA analyses.
meta_analysis.R: Perform a meta-analysis (fixed effects inverse variance weighted) on the three NL 450k batches.

analysis/c9_matches

generate_probe_sequences.R: Generate all combinations of probe sequences. Since type II probes can contain 'R' bases that represent either an A or a G (IUPAC code), there are several possible probe sequences for these probes.
get_c9matches_*: Map 450k/EPIC probes to the C9 repeat expansion using various assumptions (e.g. allowing mismatches/INDELS, assume the repeat is methylated/unmethylated, include flanking regions etc.).
process_c9matches_*: Some cleaning of the matches, includes removing duplicate matches (type I probes consist of two sequences, and type II probes of up to 8 sequences), selecting the longest match for each probe and annotating them with predicted color channels.