The GWAS pipeline - common errors¶

Here we list most common errors that might be encountered when using the pipeline.

We recommend running the pipeline with -resume argument so that progress is not lost when encountering errors.

This is a work in progress list and we will be adding more errors as people report them.

Missing 'fromPath' argument¶

This error indicates that one of the required input files was not provided by the user. Double check the list of required inputs.

Command not found¶

Check that no spaces are added at the end of arguments.

for example, if your command spans multiple lines with the use of slashes ('\') be careful to not add space after the slash as shown below.

Incorrect
--gwasmethod 'SAIGE' \

Correct
--gwasmethod 'SAIGE' \

Memory errors - termination with exit codes 2/130¶

We have set up the defaults in terms of job submission to accommodate a common set of typical usage scenarios for Genomics England data.

However, the user might get errors if they use larger files than expected or bespoke/custom resources that they created.

Command error:
  Error: Out of memory.  The --memory flag may be helpful.

To address these issues re-run the pipeline and increase the defaults for memory by using the following arguments:

--plinkmem '20000'
--memory '22 GB'
--clusterOptions '-P Bio -R rusage[mem=22000] -M 22000'

plinkmem should always be less or equal than what is provided to the processes (--memory argument) and to the cluster (--clusterOptions argument)

Check also additional optional arguments on the HPC for an even higher level of control of the memory allocated for processes (e.g., memory allocated for processes inferring the GRM).

Queue errors - unexplained termination with exit codes 130/137¶

We have set up the defaults for the submission queue for typical usage.

If your jobs unexpectedly terminate early at around 4h, then it means that the default 'short' queue is not appropriate

Change this with:

--queue 'medium'

(24h)

or

--queue 'long'

Check also additional optional arguments on the HPC for an even higher level of control of the queue for processes.

chrX has very few tested variants in the manhattan¶

siteQC for chromosome X is different. Currently we capture the info on which chromosome is processed by the first line in the .csv file of the vcf/bgen/pgen list

Make sure that chromosome is specified as "chrX" (or "chr1","chr2",...,e.t.c) in the vcf/bgen/pgen list. Not as "X" or "23".

Last update: November 3, 2023