readme.md

DNAm sites associated with ancestry

Sites will be selected by first running a GWAS of each ancestry in 1000 Genomes (https://www.internationalgenome.org/). For each ancestry, associated genetic variants that are also mQTLs (http://mqtldb.godmc.org.uk/) will be ordered by mQTL effect size and the top 50 selected for the panel.

Download and prepare GoDMC

• Input:

  • mQTL summary statistics from GoDMC http://fileserve.mrcieu.ac.uk/mqtl/assoc_meta_all.csv.gz
  • mapping between rsid and genomic coordinates ftp://hgdownload.soe.ucsc.edu/goldenPath/hg19/database/snp151Common.txt.gz

• Output: GoDMC summary statistics with hg38 coordinates in godmc-hg38.csv.gz

mkdir output
Rscript src/extract-mqtls.r godmc-hg38.csv.gz

Prepare to run ancestry GWAS

• Input: http://ftp.1000genomes.ebi.ac.uk/vol1/ftp/release/20130502/integrated_call_samples_v3.20130502.ALL.panel
• Output: GWAS phenotype files for each ancestry in pheno/

'Super-populations' (i.e. AFR AMR EAS EUR SAS) will each be compared to all other super-populations. Other populations will be compared to all other populations within the same super-population.

On the compute cluster:

1. create a folder for the scripts, inputs and outputs (in scratch space), call it BASE
2. copy ancestry/src/ to BASE
3. generate GWAS phenotype files as follows:

Rscript src/generate-phenotype-files.r pheno

Perform GWAS for each ancestry

• Input: ancestries.txt, pheno/, 1000 Genomes data
• Output: GWAS outputs for each ancestry in gwas-fst/ and gwas-glm/

On the compute cluster:

1. copy ancestry/ancestries.txt to BASE
2. download 1000 Genomes data (http://hgdownload.cse.ucsc.edu/gbdb/hg38/1000Genomes/) to BASE/1000G
3. submit the GWAS jobs to the system as follows:

mkdir gwas-glm
mkdir gwas-fst
sbatch src/gwas-glm.sh ancestries.txt 1000G pheno gwas-glm
sbatch src/gwas-fst.sh ancestries.txt 1000G pheno gwas-fst

Select top ancestry mQTLs

• Input: gwas-fst/, gwas-glm/, godmc-hg38.csv.gz, ancestries.txt
• Output: mQTLs associated with each ancestry in sites.csv (at logistic p < 5e-8)

On the compute cluster, submit the jobs to cluster as follows:

sbatch src/filter-sites.sh ancestries.txt gwas-glm gwas-fst godmc-hg38.csv.gz output/sites

Afterward, collate the top 50 for each ancestry into a single file:

Rscript src/select-sites.r output/sites output/sites.csv

Check ancestry mQTLs

Determine to what extent the SNPs identified actually capture genetic variation of ancestry.

• Input: output/sites.csv, 1000 Genomes data
• Output: Genotypes for each mQTL in output/sites.csv in VCF files in genotypes/

Submit the jobs to extract genotypes to the system as follows:

Rscript src/extract-sites.r output/sites.csv sites.txt
mkdir genotypes
sbatch src/extract-genotypes.sh sites.txt 1000G genotypes

For further checks of the CpG sites selected here, see ../checks/readme.md.