GWAS Quality Control Pipeline Project

Input:

• PLINK .bed, .bim, .fam format files (examples in /home/wheelerlab2/comp383/plink_test_data/)

Output:

• Quality control plots/reports
• Filtered PLINK .bed, .bim, .fam format files ready for GWAS
• Filtered .vcf files ready for imputation

Goal:

• Build a quality control pipeline that prepares genome-wide genotype data for downstream GWAS and other applications
• Document the process with step-by-step instructions

Useful Background

• Read the papers in this GitHub repo: https://github.com/WheelerLab/GWAS_QC/tree/master/review_articles
• Angela Andaleon’s steps: https://github.com/aandaleon/px_yri_chol
  • These steps are hardcoded for a particular dataset, we want to generalize and automate (as much as possible)

Pipeline Outline

• Most steps can be run in PLINK and results plotted with Python modules like matplotlib. We have previously used R in my lab to make plots, I would like you to use Python instead.
• Build a Python script with system calls to other tools to run the pipeline.
• Consider using a jupyter notebook to make testing plotting functions easier

1. Sex check
2. SNP call rate: plot histogram & filter
3. Person call rate: plot histogram & filter
4. Calculate Hardy-Weinberg statistics & filter SNPs
5. LD prune for relationship check & heterozygosity calculation
6. Relationship check: plot IBD stats & filter relateds (See table here for IBD info)
7. Heterozygosity check: plot & filter outliers
8. Principal component analysis (PCA) to determine genetic ancestry
   • check genome build (NCBI36/hg18 or GRCh37/hg19 or newer?)
   • automate the merge with HapMap3 genotypes (/home/wheelerlab2/Data/HAPMAP3_hg1*/)
   • run smartpca to get principal components (see documentation in /home/wheelerlab2/EIG-6.1.4/EIGENSTRAT/README)
   • plot and choose threshold for filtering people (probably can’t automate)
   • rerun smartpca with filtered set (no HapMap3)
9. Plate effects analysis (if data is available)
10. Prepare for imputation
    • make QC’d .bed, .bim, .fam files
    • use HRC or 1000G Imputation prep tool
    • make .vcf.gz files for upload to Michigan Imputation Server

Return to Course Schedule