GWIS configuration #201
GWIS configuration #201
Conversation
|
Thank you Josh, I'll have a look asap! |
| treatments = treatments_from_variant(variant, dataset) | ||
|
|
||
| if isempty(extra_treatments) | ||
| treatments = treatments_from_variant(variant, dataset) |
There was a problem hiding this comment.
In principle you could do a GWAS with 2,3,... treatments, i.e. ATEs with (variant, environmental variable) even if this is not supported yet.
Also I think this function's name is a bit ambiguous so maybe we could take the opportunity to clarify it. Something like genome_wide_estimands ? But I'm open to suggestions.
|
|
||
| if isempty(extra_treatments) | ||
| treatments = treatments_from_variant(variant, dataset) | ||
| elseif length(extra_treatments) == 1 |
There was a problem hiding this comment.
why limiting the scope? It is probably not too hard to include any order no?
| @@ -187,6 +198,24 @@ function estimands_from_gwas(dataset, variants, outcomes, confounders; | |||
| return vcat(estimands_partitions...) | |||
| end | |||
|
|
|||
| function estimands_from_gwis(dataset, variants, outcomes, confounders; | |||
There was a problem hiding this comment.
When I first read your PR with GWIS, I thought it would be whole-genome against whole genome interactions. I would personally call this mode genome wide environment interaction study, i.e. change all GWIS to GWEIS.
There was a problem hiding this comment.
This test file has no obvious difference with the GWAS mode, however I would expect the generated estimands to be quite different. Also, you can reuse a test function like get_summary_stats instead of redefining it it is exactly the same by including it in the testutils.jl file.
| elseif length(extra_treatments) == 1 | ||
| treatments = Dict(treatments_from_variant(variant, dataset)..., treatments_from_variant(string(extra_treatments[1]), dataset)...) | ||
| else | ||
| error("GWIS mode only supports pairwise interaction with one extra treatment.") |
There was a problem hiding this comment.
if someone tries to do a GWAS with extra treatment this error message will be misleading.
Hi all, I have put together a configuration for estimating pairwise interactions at the genome-wide scale. The changes to make this possible only required minimal changes to
inputs_from_config.jl. A review would be very appreciated!