This workflow implements the GEM (Gene-Environment interaction analysis for Millions of samples) tool (https://github.com/large-scale-gxe-methods/GEM). GEM conducts genome-wide gene-environment interaction tests in unrelated individuals while allowing for multiple exposures, control for genotype-covariante interactions, and robust inference.
Author: Kenny Westerman (kewesterman@mgh.harvard.edu)
GEM tool information:
- Source code: https://github.com/large-scale-gxe-methods/GEM
- Expanded documentation: https://large-scale-gxe-methods.github.io
- Manuscript: KE Westerman, DT Pham, et al. GEM: Scalable and flexible gene-environment interaction analysis in millions of samples. Bioinformatics (2021). https://doi.org/10.1093/bioinformatics/btab223.
Workflow steps:
- Run GEM (scattered across an array of input files, usually chromosomes)
- Concatenate the outputs into a single summary statistics file
Inputs:
See the "parameter_meta" section of the .wdl script.
Outputs:
- A summary statistics file containing estimates for genetic main effects and interaction effects as well as p-values for these along with a joint test of genetic main and interaction effects.
Cost estimation:
- Example analysis for reference
- Platform: Terra
- Dataset: UK Biobank (N ~ 350k)
- Analysis parameters: binary phenotype, single interaction term, 6 covariates
- Computational resources requested: 4 CPUs, 10GB memory, 250GB disk
- The above analysis completed for chromosome 2 (~1M variants) in about 9 hours, which translates to a cost of approximately $2 based on typical Terra costs assuming non-preemptible machines are used.
- To extrapolate from the above estimates: Runtime and cost should scale linearly with the sample size and number of variants and will be inversely proportional to the number of cores used.