Transform pipeline so it produces plink phenofiles and phenocode count code table.

Requirements

• Merge phewas and gwas script under plink phenofile script
• Separate phenocode from the rest of covariates and put them into a phenocode count code table
  • Research how to perform the count of the phenocode
    • Count how many times the pheno happens in a given individual vs how many times happens overall: Inspect example pheno files
• Remove and modify processes accordingly
• Update docs
• Update CI testing

Notes

For phenocode count table: As requiring multiple codes occurring on different days for a given diagnosis improves phenotype precision, users can specify count thresholds required to establish a patient as a ‘case’ for a given phenotype or simply use the code count in the regression model. From: https://academic.oup.com/bioinformatics/article/30/16/2375/2748157

Requirements

• Discuss continuity of phenotypic simulation within pipeline
• Inspect simulate pheno output
• Either:
  • Import simulation into this pipeline
  • Add wrangling of the output of the pipeline

Expose this and similarly hard-coded patterns as parameters:

Issue

Dates are not formatted accordingly during transformation

Requirements

• Add support for other date types / Improve format recognition and formatting
• Ensure everything is formatted uniformly

Issue

Requirements

• Snakecase covariates
• Improve parsing of column names -> Improve regex that snakecase them

Requirements

• Ensure query works for generating plink pheno files (regardless of phewas/gwas)
• Ensure it handles multiple hits
• Make sure that ICD10 codes are treated as they should:
  • Not used for GWAS
  • Used for pheWAS: generate code counts file

Implementation

The query functionality should read the query file from CB and perform queries accordingly to select phenotypes of interests from the different columns.

It should be able to merge or prioritise hits when multiple hits occur for a given sample in multiple measured samples

Phenotypic code (ICD, HPO, etc) should be handled in such a way that are not included in the plink phenofile, but kept apart and put into a code-count table.

s3://lifebit-featured-datasets/projects/gel/gel-gwas/

The basic idea behind this, is to pass the instances as new observations for the same individual. For this we would need to find a way to gather phenotypes per visit to the clinician.

Use case:

Assumptions
Instance 1.0 should record all phenotypes taken on XX/YY/ZZZZ date. Array 1.1, 1.2... should record multiple measurements on the same day, and they should aggregated using mean or user defined function.

Instance-array 2.0 should record all phenotypes taken on the following visit.

Solution

Instead of aggregating everything to one column, we would take all phenotypes and aggregate all the arrays using the user defined option and split them by instances as new observations for the same patient.

Support format like this: https://github.com/montilab/nf-gwas-pipeline/blob/master/data/1KG_pheno_longitudinal.csv

Issue

Vcf ids checking not needed for phenotypic data, that should be done either within the pipeline or manually from a jupyter session

Requirement

• Remove vcf ids check process and dependencies