Custom GWAS tools

Trade-offs of different methods

We provide a number of methods that enable you to run GWAS for your custom endpoint(s). This table describes the trade-offs of different methods. All methods use REGENIE. There are also instructions to use SAIGE available, but since DF7 we have moved to use REGENIE, and SAIGE has not been supported similarly since.

The fastest way to get results calculated

The easiest way to get results calculated and out of the sandbox is to use the methods in the first 2 rows - these methods enable the summary statistics to be automatically put into the green bucket and user.results pheweb browser.

If you want to browse your results outside of sandbox, the 2 lower methods require that you request to have results downloaded, which can take more time especially off-hours Helsinki time. These methods also require a bit more coding skills. Using these methods, however, you are able to modify some parameters in your job, e.g. covariates.

Category	Methods Available	Running time	Cost	When to use	How to launch	Viewing your results	Pros	Cons
Custom-GWAS	Custom GWAS Command-line version	Usually <24 hours until results in userresults.finngen.fi PheWeb	20–50€ per GWAS depending on availability in Google Cloud and increases with each DF. Charged to sandbox account	List of FinnGen IDs and case/control status in plink format (can be generated programmatically)	From terminal: finngen-cli request-gwas --help	Results automatically moved to userresults.finngen.fi (outside sandbox). Available for download and further analysis. Whole process ~24 hours	No coding required; results automatically available in PheWeb; fine-mapping can be requested from support	Cannot modify covariates (age, sex, genotyping batch, PC1–10 fixed); only releases 7 onwards
Pipelines	Command line	~8 hours for DF11	Same as above	Running >5 GWAS; need different covariates or private results; additional analysis steps (e.g., subset summary stats)	From terminal: finngen-cli request-workflow --help	View and interpret results inside sandbox using R/Python (libraries preinstalled; can request more). To view outside sandbox, download data	Can modify covariates; can run multiple phenotypes in one job	No automated import to PheWeb; requires coding skills (editing input files)
Pipelines	Standard pipelines	~8 hours for DF11	Same as above	Pipeline scalability: submit multiple phenotypes simultaneously using phenotype list	From sandbox menu: Applications > Pipelines > New job from Templates	Same as above	Same as command line pipelines	Same as command line pipelines

In the subsection below we describe the method you can use to have your results visible in userresults.finngen.fi:

• Custom GWAS commandline (CLI) tool

And in the subsections here we describe the two other methods you can use to run GWAS for your custom endpoint:

• REGENIE pipelines using command line

• How to use standard pipelines

However you can learn more about other additional methods under "How to create a user-defined endpoint".

Upstream analysis

Before entering GWAS consider investigating your phenotype and genotype cohorts. For further instructions see:

• General workflows for the most common analyses researchers are conducting with FinnGen data in FinnGen Sandbox

• Trajectory Visualization tool

• Cohort Operations tool

Downstream analysis

After custom GWAS is conducted you may consider further downstream analyses like

• How to conduct Finemapping to assign a probability that each significant variant at a locus is the causal variant. More background reading about Finemapping.

• Multiple Manhattan Plot to compare several GWAS summary statistics.