Cascading Epigenome-Wide Association Scan

This code uses the existing MetaXcan package to run a cascading analysis linking variation in genotype to traits through its effect on DNA methylation to gene expression.

Installation

Requirements

The dependencies for CEWAS can be installed via Anaconda by running the following fommand from the cewas/ directory:

conda env -f "cewas_env.yml"

The software help can then be accessed using:

conda activate cewas_env
python cewas.py -h

Testing

Included is a script entitled test_run.sh which will run CEWAS on the subset of the ROSMAP data for chromosome 22, which is included in the data directory of this repository. It will output a file named "test_results.txt" which can be compared to the file labelled "sample_output.txt" as a sanity check that the program is working as intended:

./test_run.sh
diff -d test_results.txt sample_output.txt

If nothing is printed then the sample output matches the result generated by the test.

Usage

Command help

$ python cewas.py -h
usage: cewas.py [-h] --output OUTPUT --sum_stats SUM_STATS
                [--beta_column BETA_COLUMN] [--or_column OR_COLUMN]
                [--se_column SE_COLUMN] [--pvalue_column PVALUE_COLUMN]
                [--zscore_column ZSCORE_COLUMN] --mapping MAPPING
                --reference_allele REFERENCE_ALLELE --alternative_allele
                ALTERNATIVE_ALLELE --snp_column SNP_COLUMN --cov_file_snp_fmt
                COV_FILE_SNP_FMT --cov_file_methy COV_FILE_METHY
                --predict_db_methy_fmt PREDICT_DB_METHY_FMT --predict_db_expr
                PREDICT_DB_EXPR [--ncores NCORES]

optional arguments:
  -h, --help            show this help message and exit
  --output OUTPUT       Name of file for CEWAS results (default: None)
  --sum_stats SUM_STATS
                        Path to GWAS summary statistics (default: None)
  --beta_column BETA_COLUMN
                        Name of effect size column in GWAS summary statistics
                        (default: None)
  --or_column OR_COLUMN
                        Name of odds ratio column in GWAS summary statistics
                        (default: None)
  --se_column SE_COLUMN
                        Name of standard error column in GWAS summary
                        statistics (default: None)
  --pvalue_column PVALUE_COLUMN
                        Name of P value column in GWAS summary statistics
                        (default: None)
  --zscore_column ZSCORE_COLUMN
                        Name of z-score column in GWAS summary statistics
                        (default: None)
  --mapping MAPPING     ENSEMBL mart file for re-mapping gene names (default:
                        data/ensemblHG19.tsv)
  --reference_allele REFERENCE_ALLELE
                        Reference allele column in GWAS summary statistics
                        (default: A1)
  --alternative_allele ALTERNATIVE_ALLELE
                        Alternative allele column in GWAS summary statistics
                        (default: A2)
  --snp_column SNP_COLUMN
                        SNP id column in GWAS summary statistics (default:
                        SNP)
  --cov_file_snp_fmt COV_FILE_SNP_FMT
                        Covariance file prefix for snp to methylation model
                        (default: data/covfiles/ROSMAP_CEWAS_snp_methy_cov)
  --cov_file_methy COV_FILE_METHY
                        Covariance file for methylation to expression model
                        (default:
                        data/covfiles/ROSMAP_CEWAS_pred_methy_expr_cov.txt.gz)
  --predict_db_methy_fmt PREDICT_DB_METHY_FMT
                        SNP to methylation predictdb prefix (default:
                        data/predictdb/ROSMAP_CEWAS_snp_methy)
  --predict_db_expr PREDICT_DB_EXPR
                        Methylation to expression predictdb (default:
                        data/predictdb/ROSMAP_CEWAS_methy_expr_pred.db)
  --ncores NCORES       Number of cores available for job (default: 1)

Required data

As input, CEWAS requires several models that describe the association between genotype, methylation, and gene expression in a given tissue, as well as a reference for covariance between data comprising those models. These models and covariances can be found in the data/ directory of this project.

Briefly the required data are:

• SNP to methylation predictDB, split by chromosome: This an SQLite database describing the association between genotype and DNA methylation in cis, and closely resembles the format used in MetaXcan`{name}_chr{1-22}.db"
• SNP to methylation covariance files, split by chromosome: This is a set of text files describing the covariance between SNPs in the model above, and they closely resemble the format used in MetaXcan. The files should be named as `{name}_chr{1-22}.txt.gz", and must be gzipped and whitespace delimited.
• Methylation to expression predictDB: This is an SQLite data base that describes the association between DNA methylation and Gene expression in cis, and closely resembles the format used in MetaXcan.
• Methylation to expression covariance: This is a text file describing covariance between DNA methylation probes used in the model above, and it closely resembles the format used in MetaXcan. The file must be gzipped and whitespace delimited.
• Mapping/ENSEMBL mart file: This is a text file that is used in mapping ENSEMBL IDs from the genome build corresponding to the models above (here HG19), and is used in labelling genes with their HGNC id