Skip to content

Latest commit

 

History

History
30 lines (28 loc) · 3.03 KB

index.md

File metadata and controls

30 lines (28 loc) · 3.03 KB
Raw

Full PHEASANT pipeline for UK-Biobank Nealelab GWAS.

This page describes the steps we applied in going from the raw data provided by UK-biobank, to that used to run GWAS.

  • Create a new VM on google cloud, install R and git.
  • Clone the PHESANT repository.
  • Run reengineering_phenofile_pharma_exomes.r
    • Remove cancer variables - these have a up to 31 visits and should be considered separately.
    • Restrict to the first visit (we made the assumption that the first visit contained the least missing data).
    • Prepend each column name with x and replace all instances of - and . with _, for compatibility with PHESANT.
  • Restrict to the subset of samples that are output by the genetic data QC pipeline.
  • The resultant phenotype file, restricted to the individuals for whom we have clean genetic data is then passed to PHESANT. A summary of how PHESANT parses the raw phenotype data is shown here.
  • PHESANT inputs
    • variablelistfile outcome_info_final_round2.tsv.
      • As described in the README for PHESANT, see the Variable information file heading.
        • The most important column is EXCLUDED. If there is any text in a cell in this column it is excluded.
        • CAT_MULT_INDICATOR_FIELDS is also important, it describes who to include as controls for this phenotype according to the rules explained in the README for PHESANT.
    • datacodingfile data-coding-ordinal-info.txt. See the README for PHESANT for details of its format.
    • Our default settings for filters.
      • --catmultcutoff 50. The cutoff for exclusion when creating dichotomous variables for CAT-MULTIPLE.
      • --catordnacutoff 500. The cutoff for exclusion for number of non-NAs in ordered categorical variables.
      • --catunordnacutoff 5000. The cutoff for exclusion for number of non-NAs in unordered categorical variables.
      • --contnacutoff 5000. The cutoff for exclusion for number of non-NAs in continuous variables.
      • --binnacutoff 5000. The cutoff for exclusion for number of non-NAs in binary-variables.
      • --bintruecutoff 100. The cutoff for exclusion for numbers of members of a category in binary-variables.
      • --mincategorysize 10. The minimum number of samples in a category for categorical single, integer, and continous variables.
      • --maxunorderedcategories 1000. The maximum number of categories in an unordered categorical variable.
      • --propforcontinuous 0.2. The cutoff for proportion of samples with the same value for the variable to not be considered continuous.
  • We run PHESANT in chunks on the cloud.
  • We then run the summarise_phenotypes.r functions to obtain a summary file of the phenotypes that made it into the final file.