add test for usage of 'feature_dataset_uri' instead of 'feature_generation_uri' to lazar-physchem test and to validation test

2 files changed, 67 insertions, 26 deletions

diff --git a/test/lazar-physchem-long.rb b/test/lazar-physchem-long.rb
index a4cb0b3..7959601 100644
--- a/test/lazar-physchem-long.rb
+++ b/test/lazar-physchem-long.rb
@@ -5,36 +5,70 @@ class LazarPhyschemDescriptorTest < MiniTest::Test
   def test_lazar_pc_descriptors
 
     # check available descriptors
-    desc = OpenTox::Algorithm::Descriptor.physchem_descriptors.keys
-    assert_equal 111,desc.size,"wrong num physchem descriptors"
+    @descriptors = OpenTox::Algorithm::Descriptor.physchem_descriptors.keys
+    assert_equal 111,@descriptors.size,"wrong num physchem descriptors"
     sum = 0
     {"Openbabel"=>16,"Cdk"=>50,"Joelib"=>45}.each do |k,v|
-      assert_equal v,desc.select{|x| x=~/^#{k}\./}.size,"wrong num #{k} descriptors"
+      assert_equal v,@descriptors.select{|x| x=~/^#{k}\./}.size,"wrong num #{k} descriptors"
       sum += v
     end
     assert_equal 111,sum
 
     # select descriptors for test
-    num_features_offset = 0
-    desc.keep_if{|x| x=~/^Openbabel\./}
-    desc.delete("Openbabel.L5") # TODO Openbabel.L5 does not work, investigate!!!
+    @num_features_offset = 0
+    @descriptors.keep_if{|x| x=~/^Openbabel\./}
+    @descriptors.delete("Openbabel.L5") # TODO Openbabel.L5 does not work, investigate!!!
     unless defined?($short_tests)
       # the actual descriptor calculation is rather fast, computing 3D structures takes time
       # A CDK descriptor can calculate serveral values, e.g., ALOGP produces ALOGP.ALogP, ALOGP.ALogp2, ALOGP.AMR
       # both is accepted (and tested here): Cdk.ALOGP (produces 3 features), or ALOGP.AMR (produces only 1 feature)
-      desc += ["Cdk.ALOGP.AMR", "Cdk.WienerNumbers", "Joelib.LogP", "Joelib.count.HeteroCycles"]
-      num_features_offset = 1 # Cdk.WienerNumbers produces 2 (instead of 1) features
+      @descriptors += ["Cdk.ALOGP.AMR", "Cdk.WienerNumbers", "Joelib.LogP", "Joelib.count.HeteroCycles"]
+      @num_features_offset = 1 # Cdk.WienerNumbers produces 2 (instead of 1) features
     end
-    puts "Descriptors: #{desc}"
+    puts "Descriptors: #{@descriptors}"
 
     # UPLOAD DATA
-    dataset = OpenTox::Dataset.new 
-    dataset.upload File.join(DATA_DIR,"EPAFHM.medi.csv")
-    assert_equal dataset.uri.uri?, true
-    puts "Dataset: "+dataset.uri
+    @dataset = OpenTox::Dataset.new 
+    @dataset.upload File.join(DATA_DIR,"EPAFHM.medi.csv")
+    assert_equal @dataset.uri.uri?, true
+    puts "Dataset: "+@dataset.uri
 
+    @compound_smiles = "CC(C)(C)CN"
+    @compound_inchi = "InChI=1S/C5H13N/c1-5(2,3)4-6/h4,6H2,1-3H3"
+
+    prediction_a = build_model_and_predict(true)
+    prediction_b = build_model_and_predict(false)
+    
+    assert_equal prediction_a,prediction_b,"predicted value differs depending on calculation method"
+    puts "Predicted value: #{prediction_a}"
+    # the actual value (from the complete EPAFHM dataset) is 5.45, but it is predicted higher when tested
+    # do not expect a fixed value, this might vary with, e.g., the calculated 3d structure by OB
+    assert prediction_a > 5,"predicted values should be above 5, is #{prediction_a}"
+    assert prediction_a < 15,"predicted values should be below 15, is #{prediction_a}"
+  end
+
+  def build_model_and_predict(precompute_feature_dataset=true)
+
+    model_params = {:dataset_uri => @dataset.uri}
+    feat_gen_uri = File.join($algorithm[:uri],"descriptor","physchem")
+    
+    if precompute_feature_dataset
+      # PRECOMPUTE FEATURES
+      p = "/tmp/mergedfile.csv"
+      f = File.open(p,"w")
+      f.puts File.read(File.join(DATA_DIR,"EPAFHM.medi.csv"))
+      f.puts "\"#{@compound_smiles}\","
+      f.close
+      d = OpenTox::Dataset.new 
+      d.upload p
+      model_params[:feature_dataset_uri] = OpenTox::Algorithm::Generic.new(feat_gen_uri).run({:dataset_uri => d.uri, :descriptors => @descriptors})
+    else
+      model_params[:feature_generation_uri] = feat_gen_uri
+      model_params[:descriptors] = @descriptors
+    end
+      
     # BUILD MODEL
-    model_uri = OpenTox::Model::Lazar.create :dataset_uri => dataset.uri, :feature_generation_uri => File.join($algorithm[:uri],"descriptor","physchem"), :descriptors => desc
+    model_uri = OpenTox::Model::Lazar.create model_params
     puts "Model: "+model_uri
     model = OpenTox::Model::Lazar.new model_uri
     assert_equal model_uri.uri?, true
@@ -43,9 +77,11 @@ class LazarPhyschemDescriptorTest < MiniTest::Test
     # CHECK FEATURE DATASET
     feature_dataset_uri = model.metadata[RDF::OT.featureDataset].first
     puts "Feature dataset: #{feature_dataset_uri}"
-    features = OpenTox::Dataset.new(feature_dataset_uri).features
+    feature_dataset = OpenTox::Dataset.new(feature_dataset_uri)
+    assert_equal @dataset.compounds.size,feature_dataset.compounds.size-(precompute_feature_dataset ? 1 : 0),"num compounds in feature dataset not correct"
+    features = feature_dataset.features
     feature_titles = features.collect{|f| f.title}
-    desc.each do |d|
+    @descriptors.each do |d|
       if (d=~/^Cdk\./ and d.count(".")==1) # CDK descriptors (e.g. Cdk.ALOG are included as Cdk.ALOGP.ALogP, Cdk.ALOGP.ALogp2 ..)
         match = false
         feature_titles.each do |f|
@@ -56,10 +92,10 @@ class LazarPhyschemDescriptorTest < MiniTest::Test
         assert feature_titles.include?(d),"feature not found #{d} in feature dataset #{feature_titles.inspect}"
       end
     end
-    assert_equal (desc.size+num_features_offset),features.size,"wrong num features in feature dataset"
+    assert_equal (@descriptors.size+@num_features_offset),features.size,"wrong num features in feature dataset"
 
     # predict compound
-    compound_uri = "#{$compound[:uri]}/InChI=1S/C13H8Cl2O2/c14-12-5-4-11(7-13(12)15)17-10-3-1-2-9(6-10)8-16/h1-8H"
+    compound_uri = "#{$compound[:uri]}/#{@compound_inchi}"
     prediction_uri = model.predict :compound_uri => compound_uri
     prediction = OpenTox::Dataset.new prediction_uri
     assert_equal prediction.uri.uri?, true
@@ -69,11 +105,7 @@ class LazarPhyschemDescriptorTest < MiniTest::Test
     assert prediction.features.collect{|f| f.uri}.include?(model.predicted_variable),"prediction feature #{model.predicted_variable} not included prediction dataset #{prediction.features.collect{|f| f.uri}}"
     assert prediction.compounds.collect{|c| c.uri}.include?(compound_uri),"compound #{compound_uri} not included in prediction dataset #{prediction.compounds.collect{|c| c.uri}}"
     assert_equal 1,prediction.compound_indices(compound_uri).size,"compound should only be once in the dataset"
-    predicted_value = prediction.data_entry_value(prediction.compound_indices(compound_uri).first,model.predicted_variable)
-    puts "Predicted value: #{predicted_value}"
-    assert predicted_value > 0.005,"predicted values should be above 0.005, is #{predicted_value}"
-    assert predicted_value < 0.1,"predicted values should be below 0.1, is #{predicted_value}"
-
+    prediction.data_entry_value(prediction.compound_indices(compound_uri).first,model.predicted_variable)
   end
 
 end
diff --git a/test/validation-long.rb b/test/validation-long.rb
index 88ee315..2f4fe58 100644
--- a/test/validation-long.rb
+++ b/test/validation-long.rb
@@ -65,9 +65,18 @@ class ValidationTest < MiniTest::Test
           :info => file.path, :delete => true} 
       FEAT_GEN[file].each do |feat_gen|
         data[:alg_params] = "feature_generation_uri="+feat_gen
-+       data[:alg_params] << ";backbone=false;min_chisq_significance=0.0" if feat_gen=~/fminer/ and data[:info] =~ /mini/
-        data[:alg_params] << ";descriptors="+[ "Openbabel.atoms", "Openbabel.bonds", "Openbabel.dbonds", "Openbabel.HBA1", "Openbabel.HBA2", "Openbabel.HBD", "Openbabel.MP", "Openbabel.MR", "Openbabel.MW", "Openbabel.nF", "Openbabel.sbonds", "Openbabel.tbonds", "Openbabel.TPSA"].join(",") if feat_gen=~/physchem/
-        DATA << data
+        data[:alg_params] << ";backbone=false;min_chisq_significance=0.0" if feat_gen=~/fminer/ and data[:info] =~ /mini/
+        if feat_gen=~/physchem/
+          # validation with physchem descriptors is performed twice, once with feature_generation_uri, once with feature_dataset_uri
+          desc = [ "Openbabel.atoms", "Openbabel.bonds", "Openbabel.dbonds", "Openbabel.HBA1", "Openbabel.HBA2", "Openbabel.HBD", "Openbabel.MP", "Openbabel.MR", "Openbabel.MW", "Openbabel.nF", "Openbabel.sbonds", "Openbabel.tbonds", "Openbabel.TPSA"]
+          data[:alg_params] << ";descriptors="+desc.join(",")
+          DATA << data
+          feature_dataset_uri = OpenTox::Algorithm::Generic.new(feat_gen).run({:dataset_uri => data[:data], :descriptors => desc})
+          data[:alg_params] = "feature_dataset_uri="+feature_dataset_uri
+          DATA << data
+        else
+          DATA << data
+        end
       end
     end
   end