UK Biobank GxE

This project explores GxE effects in the UK biobank.

Phenotype-covariate table preparation

The phenotype-covariate table is pre-processed in the UKB RAP with the following filters:

• Outliers for heterozygosity or missing rate (field 22027) IS NOT "Yes"
• Genetic ethnic grouping (field 22006) IS "Caucasian"
• Smoking status | Instance 0 (field 20116.0) IS NOT "-3" (Prefer not to answer)
• Body mass index (BMI) | Instance 0 (field 21001.0) IS NOT NULL

I always chose instance 0 of the measurements (first assessment at recruitment) because is the one where most measures are available. The filtered table is converted to csv using the Table exporter application (https://ukbiobank.dnanexus.com/app/app-GYXYYzQ9j5bFJ8x618yyb2jf).

Other sample filters

I drop close relatives (up to 3rd degree). The list of dropped IDs is generated with the notebook ./filter_close_relatives.ipynp.

Genotype preparation

I use the WTCHG imputed genotypes (field 22828). I convert them to pgen format using ./filter_bgen.sh. During the conversion I also:

• Drop variants with MAF smaller than 0.01
• Drop samples dropped from the phenotype-covariate table or because they have close relatives (notebook ./get_final_id_list.ipynb)
• Sort samples in the same order as the phenotype-covariate table