diff options
| -rw-r--r-- | .gitignore | 1 | ||||
| -rw-r--r-- | config.ru | 2 | ||||
| -rw-r--r-- | lazar.rb | 262 |
3 files changed, 155 insertions, 110 deletions
@@ -2,3 +2,4 @@ tmp/* *.sqlite3 log/* db/* +models/* @@ -4,7 +4,7 @@ require 'application.rb' require 'rack' require 'rack/contrib' -['db','public','log','tmp'].each do |dir| +['models','public','log','tmp'].each do |dir| FileUtils.mkdir_p dir unless File.exists?(dir) end @@ -1,134 +1,178 @@ -require 'datamapper' +require 'redland' +require 'rdf/redland' +require 'rdf/redland/util' -DataMapper::setup(:default, "sqlite3://#{Dir.pwd}/db/#{ENV['RACK_ENV']}.db") +@@storage = Redland::MemoryStore.new +@@parser = Redland::Parser.new +@@serializer = Redland::Serializer.new -class LazarModel - include DataMapper::Resource - property :id, Serial - property :activity_dataset_uri, String, :length => 256 # default is too short for URIs - property :feature_dataset_uri, String, :length => 256 # default is too short for URIs - property :created_at, DateTime +get '/?' do # get index of models + Dir["models/*"].collect{|model| url_for("/", :full) + File.basename(model,".yaml")}.sort.join("\n") +end - def uri - File.join(OpenTox::Model::LazarClassification.base_uri,"lazar_classification", self.id.to_s) +get '/:id/?' do + path = File.join("models",params[:id] + ".yaml") + if File.exists? path + send_file path + else + status 404 + "Model #{params[:id]} does not exist." end +end - def predict(compound) - - training_activities = OpenTox::Dataset.find :uri => @activity_dataset_uri - # TODO: find database activities - # TODO: find prediction - training_features = OpenTox::Dataset.find :uri => @feature_dataset_uri - - prediction_dataset = OpenTox::Dataset.find_or_create(:name => training_activities.name + '_predictions') - prediction_neighbors = OpenTox::Dataset.find_or_create(:name => training_activities.name + '_neighbors') - prediction_features = OpenTox::Dataset.find_or_create(:name => training_activities.name + '_prediction_features') - - feature_uris = compound.match(training_features) - prediction_features.add({compound.uri => feature_uris}.to_yaml) - - conf = 0.0 - neighbors = [] - - training_features.compounds.each do |neighbor| - sim = OpenTox::Algorithm::Similarity.weighted_tanimoto(training_features,neighbor,prediction_features,compound).to_f - if sim > 0.3 - neighbors << neighbor.uri - training_activities.features(neighbor).each do |a| - case OpenTox::Feature.new(:uri => a.uri).value('classification').to_s - when 'true' - conf += OpenTox::Utils.gauss(sim) - when 'false' - conf -= OpenTox::Utils.gauss(sim) - end - end - end +delete '/:id/?' do + path = File.join("models",params[:id] + ".yaml") + if File.exists? path + File.delete path + "Model #{params[:id]} deleted." + else + status 404 + "Model #{params[:id]} does not exist." + end +end + +post '/?' do # create model + halt 404, "Dataset #{params[:activity_dataset_uri]} not found" unless OpenTox::Dataset.find(params[:activity_dataset_uri]) + halt 404, "Dataset #{params[:feature_dataset_uri]} not found" unless OpenTox::Dataset.find(params[:feature_dataset_uri]) + activities = Redland::Model.new @storage + features = Redland::Model.new @storage + training_activities = OpenTox::Dataset.find params[:activity_dataset_uri] + training_features = OpenTox::Dataset.find params[:feature_dataset_uri] + @@parser.parse_string_into_model(activities,training_activities,'/') + @@parser.parse_string_into_model(features,training_features,'/') + feature = Redland::Node.new(Redland::Uri.new(File.join(@@config[:services]["opentox-algorithm"],'fminer'))) + p_value = Redland::Node.new(Redland::Uri.new(File.join(@@config[:services]["opentox-algorithm"],'fminer/p_value'))) + effect = Redland::Node.new(Redland::Uri.new(File.join(@@config[:services]["opentox-algorithm"],'fminer/effect'))) + + smarts = [] + p_vals = {} + effects = {} + fingerprints = {} + features.triples do |s,p,o| + s = s.uri.to_s.sub(/^\//,'') + case p + when feature + fingerprints[s] = [] unless fingerprints[s] + fingerprints[s] << o.uri.to_s.sub(/^\//,'') + when p_value + sma = s.to_s + smarts << sma + p_vals[sma] = o.to_s.to_f + when effect + sma = s.to_s + effects[sma] = o.to_s end - conf = conf/neighbors.size - if conf > 0.0 - classification = true - elsif conf < 0.0 - classification = false + end + + activity_uris = [] + act = {} + activities.triples do |s,p,o| + activity_uris << p.uri.to_s + s = s.uri.to_s + case o.to_s + when "true" + act[s] = true + when "false" + act[s] = false end + end - prediction = OpenTox::Feature.new(:name => training_activities.name + " prediction", :classification => classification, :confidence => conf) - prediction_neighbors.add({compound.uri => neighbors}.to_yaml) - prediction_dataset.add({compound.uri => [prediction.uri]}.to_yaml) + activity_uris.uniq! + if activity_uris.size != 1 + halt 400 + "Dataset #{params[:activity_dataset_uri]} has not exactly one feature." + end - prediction.uri + id = Dir["models/*"].collect{|models| File.basename(models,".yaml").to_i}.sort.last + if id.nil? + id = 1 + else + id += 1 + end + File.open(File.join("models",id.to_s + ".yaml"),"w") do |f| + f.write({ + :endpoint => activity_uris[0], + :features => smarts, + :p_values => p_vals, + :effects => effects, + :fingerprints => fingerprints, + :activities => act + }.to_yaml) end + url_for("/#{id}", :full) + end -# automatically create the post table -LazarModel.auto_migrate! #unless LazarModel.table_exists? -LazarModel.auto_migrate! if ENV['RACK_ENV'] == 'test' +# PREDICTIONS +post '/:id/?' do # create prediction + path = File.join("models",params[:id] + ".yaml") + if !File.exists? path + status 404 + "Model #{params[:id]} does not exist." + end -get '/lazar_classification/?' do # get index of models - LazarModel.all.collect{|m| m.uri}.join("\n") -end + compound = OpenTox::Compound.new :uri => params[:compound_uri] -get '/lazar_classification/:id/?' do - halt 404, "Model #{params[:id]} not found." unless @model = LazarModel.get(params[:id]) - @model.to_yaml -end + data = YAML.load_file path -delete '/lazar_classification/:id/?' do - halt 404, "Model #{params[:id]} not found." unless @model = LazarModel.get(params[:id]) - @model.destroy - "Model #{params[:id]} succesfully deleted." -end + # find database activities + if data[:activities][compound.uri] + output = Redland::Model.new @storage + output.add Redland::Uri.new(compound.uri), Redland::Uri.new(data[:endpoint]), Redland::Literal.new(data[:activities][compound.uri].to_s) + halt 200, @@serializer.model_to_string(Redland::Uri.new(url_for("/",:full)), output) + end -post '/lazar_classification/?' do # create model - halt 404, "Dataset #{params[:activity_dataset_uri]} not found" unless OpenTox::Dataset.find(:uri => params[:activity_dataset_uri]) - halt 404, "Dataset #{params[:feature_dataset_uri]} not found" unless OpenTox::Dataset.find(:uri => params[:feature_dataset_uri]) - model = LazarModel.new(params) - model.save - model.uri -end + compound_matches = compound.match data[:features] -# PREDICTIONS -post '/lazar_classification/:id/?' do # create prediction - halt 404, "Model #{params[:id]} not found." unless @model = LazarModel.get(params[:id]) - compound = OpenTox::Compound.new :uri => params[:compound_uri] - @model.predict(compound) -end + conf = 0.0 + neighbors = [] + classification = nil -get '/lazar_classification/*/predictions?' do # get dataset URI - name = params[:splat].first - halt 404, "Model #{name} not found." unless @model = LazarModel.get(request.url) - # Dataset.find -end + data[:fingerprints].each do |uri,matches| + sim = weighted_tanimoto(compound_matches,matches,data[:p_values]) + if sim > 0.3 -get '/lazar_classification/*/prediction/*' do # display prediction for a compound - name = params[:splat].first - compound_uri = params[:splat][1] - halt 404, "Model #{name} not found." unless @model = LazarModel.get(request.url) - # prediction not found - #prediction.to_yaml - #xml prediction -end + neighbors << uri + case data[:activities][uri].to_s + when 'true' + conf += OpenTox::Utils.gauss(sim) + when 'false' + conf -= OpenTox::Utils.gauss(sim) + end + end + conf = conf/neighbors.size + if conf > 0.0 + classification = true + elsif conf < 0.0 + classification = false + end + end -get '/lazar_classification/*/prediction/*/neighbors' do - name = params[:splat].first - compound_uri = params[:splat][1] - halt 404, "Model #{name} not found." unless @model = LazarModel.get(request.url) - # prediction not found - # prediction.neighbors + output = Redland::Model.new @storage + + output.add Redland::Uri.new(compound.uri), Redland::Uri.new(url_for("/#{params[:id]}/classification",:full)), classification.to_s + output.add Redland::Uri.new(compound.uri), Redland::Uri.new(url_for("/#{params[:id]}/confidence",:full)), conf.to_s + @@serializer.model_to_string(Redland::Uri.new(url_for("/",:full)), output) + +# { :classification => classification, +# :confidence => conf, +# :neighbors => neighbors, +# :features => compound_matches +# }.to_yaml end -get '/lazar_classification/*/prediction/*/features' do - name = params[:splat].first - compound_uri = params[:splat][1] - halt 404, "Model #{name} not found." unless @model = LazarModel.get(request.url) - # prediction not found - # prediction not found - # prediction.features -end -delete '/lazar_classification/*/prediction/*' do # display prediction for a compound - name = params[:splat].first - halt 404, "Model #{name} not found." unless @model = LazarModel.get(request.url) - # Prediction.destroy +def weighted_tanimoto(fp_a,fp_b,p) + common_features = fp_a & fp_b + all_features = fp_a + fp_b + common_p_sum = 0.0 + if common_features.size > 0 + common_features.each{|f| common_p_sum += p[f]} + all_p_sum = 0.0 + all_features.each{|f| all_p_sum += p[f]} + common_p_sum/all_p_sum + else + 0.0 + end end - |