Changes from GWIS to GWIES; Updated tests and generalized estimand creation to support higher order interactions

Author: joshua-slaughter

Manifest.toml

@@ -2,7 +2,7 @@
 
 julia_version = "1.10.4"
 manifest_format = "2.0"
-project_hash = "b1a916e4f68d2fb953c6eba21c0fc7546e7dc5c6"
+project_hash = "4182f1a89c1532988a40d605eeec69822a2af3fc"
 
 [[deps.ARFFFiles]]
 deps = ["CategoricalArrays", "Dates", "Parsers", "Tables"]
@@ -1358,9 +1358,9 @@ weakdeps = ["ChainRulesCore", "SparseArrays", "Statistics"]
 
 [[deps.LogExpFunctions]]
 deps = ["DocStringExtensions", "IrrationalConstants", "LinearAlgebra"]
-git-tree-sha1 = "a2d09619db4e765091ee5c6ffe8872849de0feea"
+git-tree-sha1 = "13ca9e2586b89836fd20cccf56e57e2b9ae7f38f"
 uuid = "2ab3a3ac-af41-5b50-aa03-7779005ae688"
-version = "0.3.28"
+version = "0.3.29"
 
     [deps.LogExpFunctions.extensions]
     LogExpFunctionsChainRulesCoreExt = "ChainRulesCore"

Project.toml

@@ -15,6 +15,7 @@ DataFrames = "a93c6f00-e57d-5684-b7b6-d8193f3e46c0"
 Distributions = "31c24e10-a181-5473-b8eb-7969acd0382f"
 HypothesisTests = "09f84164-cd44-5f33-b23f-e6b0d136a0d5"
 JLD2 = "033835bb-8acc-5ee8-8aae-3f567f8a3819"
+LogExpFunctions = "2ab3a3ac-af41-5b50-aa03-7779005ae688"
 MKL = "33e6dc65-8f57-5167-99aa-e5a354878fb2"
 Mmap = "a63ad114-7e13-5084-954f-fe012c677804"
 OrderedCollections = "bac558e1-5e72-5ebc-8fee-abe8a469f55d"
@@ -37,12 +38,12 @@ CairoMakie = "0.12"
 CategoricalArrays = "0.10"
 Combinatorics = "1.0"
 DataFrames = "1.2"
+MKL = "0.7"
 OrderedCollections = "1.6.3"
 PackageCompiler = "2.1.17"
 SnpArrays = "0.3"
 StableRNGs = "1.0.1"
 Statistics = "1.10"
 TMLE = "0.17"
-MKL = "0.7"
 YAML = "0.4.9"
 julia = "1.10, 1"

src/inputs_from_config.jl

@@ -54,7 +54,7 @@ function try_append_new_estimands!(
     end
 end
 
-function estimands_from_variants(
+function genome_wide_estimands(
     variants,
     dataset, 
     estimand_constructor, 
@@ -70,10 +70,11 @@ function estimands_from_variants(
 
         if isempty(extra_treatments)
             treatments = treatments_from_variant(variant, dataset)
-        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.")
+            treatments = Dict(
+                treatments_from_variant(variant, dataset)..., 
+                (pair for extra in extra_treatments for pair in treatments_from_variant(string(extra), dataset))...
+            )
         end
 
         local Ψ
@@ -172,7 +173,7 @@ end
 """
     treatment_from_variant(variant, dataset)
 
-    Generate a key-value pair (dicitionary) for treatment structs.
+    Generate a key-value pair (dictionary) for treatment structs.
 """
 function treatments_from_variant(variant::String, dataset::DataFrame)
     variant_levels = sort(levels(dataset[!, variant], skipmissing=true))
@@ -187,7 +188,7 @@ function estimands_from_gwas(dataset, variants, outcomes, confounders;
     )
     variants_groups = Iterators.partition(variants, length(variants) ÷ Threads.nthreads())
     estimands_tasks = map(variants_groups) do variants
-        Threads.@spawn estimands_from_variants(variants, dataset, ATE, outcomes, confounders;
+        Threads.@spawn genome_wide_estimands(variants, dataset, ATE, outcomes, confounders;
             extra_treatments=extra_treatments,
             outcome_extra_covariates=outcome_extra_covariates,
             positivity_constraint=positivity_constraint,
@@ -198,15 +199,15 @@ function estimands_from_gwas(dataset, variants, outcomes, confounders;
     return vcat(estimands_partitions...)
 end
 
-function estimands_from_gwis(dataset, variants, outcomes, confounders;
+function estimands_from_gweis(dataset, variants, outcomes, confounders;
     extra_treatments=extra_treatments,
     outcome_extra_covariates = [],
     positivity_constraint=0.,
     verbosity=0
     )
     variants_groups = Iterators.partition(variants, length(variants) ÷ Threads.nthreads())
     estimands_tasks = map(variants_groups) do variants
-        Threads.@spawn estimands_from_variants(variants, dataset, AIE, outcomes, confounders;
+        Threads.@spawn genome_wide_estimands(variants, dataset, AIE, outcomes, confounders;
             extra_treatments=extra_treatments,
             outcome_extra_covariates=outcome_extra_covariates,
             positivity_constraint=positivity_constraint,
@@ -239,7 +240,7 @@ function get_genotypes_from_beds(bedprefix)
 end
 
 function make_genotypes(genotype_prefix, config, call_threshold)
-    genotypes = if config["type"] == "gwas"|| config["type"] == "gwis"
+    genotypes = if config["type"] == "gwas"|| config["type"] == "gweis"
         get_genotypes_from_beds(genotype_prefix)
     else
         variants_set = Set(retrieve_variants_list(config["variants"]))
@@ -297,9 +298,9 @@ function inputs_from_config(config_file, genotypes_prefix, traits_file, pcs_file
             positivity_constraint=positivity_constraint,
             verbosity=verbosity
         )
-    elseif config_type == "gwis"
+    elseif config_type == "gweis"
         variants = filter(!=("SAMPLE_ID"), names(genotypes))
-        estimands_from_gwis(dataset, variants, outcomes, confounders;
+        estimands_from_gweis(dataset, variants, outcomes, confounders;
             extra_treatments=extra_treatments,
             outcome_extra_covariates=outcome_extra_covariates,
             positivity_constraint=positivity_constraint,

test/data/config_gwis.yaml

@@ -10,6 +10,6 @@ TESTDIR = joinpath(pkgdir(TargeneCore), "test")
     @test include(joinpath(TESTDIR, "inputs_from_estimands.jl"))
     @test include(joinpath(TESTDIR, "inputs_from_config.jl"))
     @test include(joinpath(TESTDIR, "inputs_from_gwas_config.jl"))
-    @test include(joinpath(TESTDIR, "inputs_from_gwis_config.jl"))
+    @test include(joinpath(TESTDIR, "inputs_from_gweis_config.jl"))
     @test include(joinpath(TESTDIR, "sieve_variance.jl"))
 end