GWAS

Inputs

This section describes the main inputs to the workflow, see further down for the associated Nextflow parameters.

Genetic Data

For this workflow, both genotypes and imputed genotypes are required. They will likely be the respective outputs of the Combining Datasets and Genotypes Imputation workflows, optionally merged with Combining with UK Biobank depending on your project of interest.

### Covariates

There are two main types of covariates files you can use here.

• The COVARIATES file contains the usual covariates information including phenotypes and is typically provided by Dominique on a per project basis. Unfortunately, at this point, the specifications for this file have not been clearly defined. The columns of the file I have received are: genotype_file_id, age_years, sex, case_or_control, cohort, severe_cohort_primary_diagnosis, isaric_cohort_max_severity_score.
• An optional INFERRED_COVARIATES contains covariates information that is inferred from the genetic datasets during the Combining Datasets workflow. This is the case for ancestry estimates for instance.

GWAS Variables

You can define the phenotypes, covariates and groups via an external YAML file which looks like the following.

groupby: ["ANCESTRY_ESTIMATE"]
phenotypes: ["SEVERE_COVID_19"]
covariates: ["AGE", "SEX", "AGE_x_SEX", "AGE_x_AGE"]

In this example:

• An independent GWAS will be run for each ancestry group in the ANCESTRY_ESTIMATE column.
• There is only one phenotype: SEVERE_COVID_19.
• The covariates to are $AGE$, $SEX$, $AGE^2$ and $AGE \cdot AGE$.

Groups

The groupby variables defines groups for which an independent GWAS will be run. If the $groupby$ key is not present in the YAML file, only one GWAS will be run for the whole population.

Phenotypes

The phenotype variables must either be case_or_control or have an existing definition based on other columns in the COVARIATES file (hardcoded), at the moment only SEVERE_COVID_19 has been defined.

Covariates

The covariates variables are used as adjustment variables in the regression model. In REGENIE step 2, principal components are also added in a leave-one-chromosome-out scheme to the covariates defined here.

Again, due to the ill-definition of the covariates file, at the moment, only a subset of non-inferred covariates can be processed: AGE and SEX. Categorical covariates are one-hot encoded.

Workflow Parameters

This is the list of all the pipeline's parameters, they can be set in the run.config file under the params section.

• GENOTYPES_PREFIX: Prefix to genotypes in PLINK BED format (likely the output of the Combining Datasets workflow).
• IMPUTED_GENOTYPES_PREFIX: Prefix to imputed genotypes in PGEN format (likely the output of the Genotypes Imputation workflow).
• COVARIATES: Path to covariate file (likely the output of the Combining Datasets workflow)
• INFERRED_COVARIATES: Optional, path to covariates inferred from genetic data during the Combining Datasets workflow.
• N_PCS (default: 10): Number of principal components to compute.
• PCA_APPROX (default: true): Whether PCA is performed via approximation see
• MIN_GROUP_SIZE (default: 100): Minimum number of samples in a group to proceed to effect size estimation.
• VARIABLES_CONFIG (default: assets/variables.yaml): File containing the declaration of groups, phenotypes and covariates for the GWAS (see GWAS Variables).
• REGENIE_MAF (default: 0.01): Minor allele frequency for a variant to enter the GWAS.
• REGENIE_MAC (default: 10): Minor allele count for a variant to enter the GWAS.
• REGENIE_BSIZE (default: 1000): Regenie's block size (see the regenie docs)

Running The Workflow

If the previous steps have been completed successfully you can run on your platform (e.g. PLATFORM=eddie):

nextflow run main.nf -entry GWAS -resume -profile PLATFORM -with-trace -with-report -c run.config

Outputs

All outputs are produced in PUBLISH_DIR (default: results), for each group a separate folder contains:

• plots: Manhattan and QQ plots
• results: A CSV file of summary statistics

Workflow DAG