Test for interactions with string-treatments and integer extra treatments

Author: joshua-slaughter

src/inputs_from_config.jl

@@ -193,7 +193,13 @@ end
 """
 function treatments_from_variant(variant::String, dataset::DataFrame)
     variant_levels = sort(levels(dataset[!, variant], skipmissing=true))
-    return Dict{Symbol, Vector{String}}(Symbol(variant)=>variant_levels)
+    if typeof(variant_levels) == Vector{String}
+        return Dict{Symbol, Vector{String}}(Symbol(variant)=>variant_levels)
+    elseif typeof(variant_levels) == Vector{Int}
+        return Dict{Symbol, Vector{UInt8}}(Symbol(variant)=>variant_levels)
+    else
+        throw(ArgumentError("Unknown type for treatment levels (must be of type String or Int)"))
+    end
 end
 
 function estimands_from_gwas(dataset, variants, outcomes, confounders; 

test/inputs_from_gweis_config.jl

@@ -20,7 +20,7 @@ function get_summary_stats(estimands)
     return sort(combine(groupby(results, :OUTCOME), nrow), :OUTCOME)
 end
 
-function check_estimands_levels_interactions(estimands)
+function check_estimands_levels_interactions(estimands, snp_info)
     extra_treatments = YAML.load_file(joinpath(TESTDIR, "data", "config_gweis_first_order.yaml"))["extra_treatments"]
     for (i,x) in enumerate(extra_treatments)
         extra_treatments[i]=Symbol(x)
@@ -31,18 +31,27 @@ function check_estimands_levels_interactions(estimands)
         # The variant should always be the last key
         treatment_set = collect(keys(Ψ.args[1].treatment_values))
         variant = setdiff(treatment_set, extra_treatments)[1]
+        variant_info = filter(:snpid=>x->x==String(variant),snp_info)
+        allele1, allele2 = variant_info.allele1[1], variant_info.allele2[1]
+        # Here, we check if the order is sufficient to be able to compute non-linear effects any of these combinations will do
         if length(Ψ.args) == 2
-            @test Ψ.args[1].treatment_values[variant] == (control = 0x00, case = 0x01)
-            @test Ψ.args[2].treatment_values[variant] == (control = 0x01, case = 0x02)
+            @test (Ψ.args[1].treatment_values[variant] == (control = allele1*allele1, case = allele1*allele2) && 
+            Ψ.args[2].treatment_values[variant] == (control = allele1*allele2, case = allele2*allele2)) || 
+            (Ψ.args[1].treatment_values[variant] == (control = allele2*allele2, case = allele1*allele2) && 
+            Ψ.args[2].treatment_values[variant] == (control = allele1*allele2, case = allele1*allele1))
         else
             # Otherwise we check they are one or the other
             arg = only(Ψ.args)
-            @test arg.treatment_values[variant]==(control = 0x00, case = 0x01) ||
-            arg.treatment_values[variant]==( control = 0x01, case = 0x02)
+            @test arg.treatment_values[variant] == (control = allele1*allele1, case = allele1*allele2)  ||
+            arg.treatment_values[variant] == (control = allele2*allele2, case = allele1*allele2) ||
+            arg.treatment_values[variant] == (control = allele1*allele2, case = allele2*allele2) ||
+            arg.treatment_values[variant] == (control = allele1*allele2, case = allele1*allele1) 
+
         end
    end
 end
 
+
 @testset "Test inputs_from_config gweis: no positivity constraint" begin
     tmpdir = mktempdir()
     copy!(ARGS, [
@@ -76,8 +85,11 @@ end
         OUTCOME = [:BINARY_1, :BINARY_2, :CONTINUOUS_1, :CONTINUOUS_2], 
         nrow = repeat([2625], 4)
     )
-    println(estimands[1])
-    check_estimands_levels_interactions(estimands)
+    # Define SNP information to check string allele defintions
+    snpdata = read_bed_chromosome(joinpath(TESTDIR, "data", "ukbb", "genotypes" , "ukbb_1."))
+    snp_info = select(DataFrame(snpdata.snp_info), [:snpid, :allele1, :allele2])
+
+    check_estimands_levels_interactions(estimands, snp_info)
 end
 
 
@@ -113,7 +125,11 @@ end
         nrow = repeat([430], 4)
     )
 
-    check_estimands_levels_interactions(estimands)
+     # Define SNP information to check string allele defintions
+     snpdata = read_bed_chromosome(joinpath(TESTDIR, "data", "ukbb", "genotypes" , "ukbb_1."))
+     snp_info = select(DataFrame(snpdata.snp_info), [:snpid, :allele1, :allele2])
+
+    check_estimands_levels_interactions(estimands, snp_info)
 end