gwas-scripts

A repository containing various scripts useful for performing quality control on data from genome-wide association studies and visualizing results

Scripts contained in this repository:

(1) qqplot.R

A script for generating a standard quantile-quantile (QQ) plot and calculating lambda for your data. The script takes in an input file containing the column "P" (with your GWAS p-values).

The script is run as follows:

Rscript qqplot.R inputfile stat_type output_file color xmamx

  stat_type: P (for p-values), Z (for z-scores), or CHISQ (for chi-square values)
  color: color you want your points plotted in
  xmax: maximum value for the x-axis (and y-axis, as the plot is square)

(2) qq_by_maf.R

A script for generating a QQ plot that plots data stratified by allele frequency. The input data must contain the following columns: P (p-values from your GWAS), FRQ (the frequency of the tested allele from your GWAS).

The script runs as follows:

Rscript qq_by_maf.R inputfile outputfile

(3) qq_by_info.R

A script for generating a QQ plot that plots data stratified by imputation quality. Does exactly the same as the QQ by maf script, but stratifying by imputation quality (INFO) score (typically a number between 0 and 1). The input data must contain the following columns: P (p-values from your GWAS), INFO (imputation quality score).

The script runs as follows:

Rscript qq_by_info.R inputfile outputfile

(4) manhattan.R

A script for plotting a Manhattan plot. Note that this depends on having the "qqman" library (written by Stephen Turner; please see his excellent GitHub for more information https://github.com/stephenturner/qqman). The input data must contain the following columns and be in this order: SNP (rsID), CHR (chromosome), BP (base position), P (p-value)

(5) qq.three_plots.R

Combines scripts 1-3 into a single script to produce three QQ plots at once. The columns of your data must include: P, FRQ, and INFO.

The script runs as follows:

Rscript qqplot.R inputfile stat_type output_file color xmamx

  stat_type: P (for p-values), Z (for z-scores), or CHISQ (for chi-square values)
  color: color you want your points plotted in
  xmax: maximum value for the x-axis (and y-axis, as the plot is square)

(6) lambda.R

A quick script to calculate lambda in your data. Your data must have column "P" containing p-values.

The script runs as follows:

Rscript lambda.R inputfile stat_type

stat_type: P (for p-values), Z (for z-scores), or CHISQ (for chi-square values)

(7) prep.PCA.sh

A script for preparing (Plink-formatted) data to run principal component analysis.

(8) run_smartPCA.project.sh

A script that makes a call to EIGENSTRAT to plot
(a) your GWAS samples, projected onto
  (b) a reference dataset (for example, HapMap 3)

Input data should be:
(a) A Plink-formatted PED file containing the genotypes from your GWAS samples and the samples from your reference data. (b) A Plink-formatted MAP file, corresponding with the PED file, containing marker information. (c) A PEDIND file. Contains the first 5 columns of the PED file. The 6th column indicates "test" (i.e., a sample from your GWAS data) or "ref" (i.e., a sample from your reference data.

The script runs as follows:

./run_smartPCA.project.sh file-root-name

where file-root-name is the root name of your PED, MAP and PEDIND files.

(9) run_smartPCA.sh

Exactly as above to run principal component analysis in EIGENSTRAT, but only performing the analysis within your own GWAS data samples. You therefore only need a PED and MAP file.

The script runs as follows:

./run_smartPCA.sh file-root-name

where file-root-name is the root name of your PED and MAP files.