diff options
-rwxr-xr-x | lazar.rb | 145 |
1 files changed, 130 insertions, 15 deletions
@@ -1,8 +1,120 @@ +# R integration +# workaround to initialize R non-interactively (former rinruby versions did this by default) +R = nil +require ("rinruby") # this requires R to be built with X11 support (implies package xorg-dev) not longer true with this hack (ch) +@@r = RinRuby.new(false,false) +@@r.eval "library('kernlab')" # this requires R package "kernlab" to be installed + class Lazar < Model attr_accessor :prediction_dataset - def classify(compound_uri,prediction) + # AM begin + # regression function, created 06/10 + # ch: please properly integrate this into the workflow. You will need some criterium for distinguishing regression/classification (hardcoded regression for testing) + def regression(compound_uri,prediction) + + lazar = YAML.load self.yaml + compound = OpenTox::Compound.new(:uri => compound_uri) + + # obtain X values for query compound + compound_matches = compound.match lazar.features + + conf = 0.0 + similarities = {} + regression = nil + + regr_occurrences = [] # occurrence vector with {0,1} entries + sims = [] # similarity values between query and neighbors + acts = [] # activities of neighbors for supervised learning + neighbor_matches = [] # as in classification: URIs of matches + gram_matrix = [] # square matrix of similarities between neighbors; implements weighted tanimoto kernel + i = 0 + + # aquire data related to query structure + lazar.fingerprints.each do |uri,matches| + sim = OpenTox::Algorithm::Similarity.weighted_tanimoto(compound_matches,matches,lazar.p_values) + lazar.activities[uri].each do |act| + if sim > 0.3 + similarities[uri] = sim + conf += OpenTox::Utils.gauss(sim) + sims << OpenTox::Utils.gauss(sim) + #TODO check for 0 s + acts << Math.log10(act.to_f) + #acts << act.to_f + neighbor_matches[i] = matches + i+=1 + end + end + end + conf = conf/similarities.size + LOGGER.debug "Regression: found " + neighbor_matches.size.to_s + " neighbors." + + + unless neighbor_matches.length == 0 + # gram matrix + (0..(neighbor_matches.length-1)).each do |i| + gram_matrix[i] = [] + # lower triangle + (0..(i-1)).each do |j| + sim = OpenTox::Algorithm::Similarity.weighted_tanimoto(neighbor_matches[i], neighbor_matches[j], lazar.p_values) + gram_matrix[i] << OpenTox::Utils.gauss(sim) + end + # diagonal element + gram_matrix[i][i] = 1.0 + # upper triangle + ((i+1)..(neighbor_matches.length-1)).each do |j| + sim = OpenTox::Algorithm::Similarity.weighted_tanimoto(neighbor_matches[i], neighbor_matches[j], lazar.p_values) + gram_matrix[i] << OpenTox::Utils.gauss(sim) + end + end + + LOGGER.debug "Setting R data ..." + # set data + @@r.gram_matrix = gram_matrix.flatten + @@r.n = neighbor_matches.length + @@r.y = acts + @@r.sims = sims + + LOGGER.debug "Preparing R data ..." + # prepare data + @@r.eval "y<-as.vector(y)" + @@r.eval "gram_matrix<-as.kernelMatrix(matrix(gram_matrix,n,n))" + @@r.eval "sims<-as.vector(sims)" + + # model + support vectors + LOGGER.debug "Creating SVM model ..." + @@r.eval "model<-ksvm(gram_matrix, y, kernel=matrix, type=\"nu-svr\", nu=0.8)" + @@r.eval "sv<-as.vector(SVindex(model))" + @@r.eval "sims<-sims[sv]" + @@r.eval "sims<-as.kernelMatrix(matrix(sims,1))" + LOGGER.debug "Predicting ..." + @@r.eval "p<-predict(model,sims)[1,1]" + regression = 10**(@@r.p.to_f) + LOGGER.debug "Prediction is: '" + regression.to_s + "'." + + end + + if (regression != nil) + feature_uri = lazar.dependentVariables + prediction.compounds << compound_uri + prediction.features << feature_uri + prediction.data[compound_uri] = [] unless prediction.data[compound_uri] + tuple = { + File.join(@@config[:services]["opentox-model"],"lazar#regression") => regression, + File.join(@@config[:services]["opentox-model"],"lazar#confidence") => conf, + File.join(@@config[:services]["opentox-model"],"lazar#similarities") => similarities, + File.join(@@config[:services]["opentox-model"],"lazar#features") => compound_matches + } + prediction.data[compound_uri] << {feature_uri => tuple} + end + + + end + # AM end + + + def classification(compound_uri,prediction) lazar = YAML.load self.yaml compound = OpenTox::Compound.new(:uri => compound_uri) @@ -40,15 +152,10 @@ class Lazar < Model prediction.features << feature_uri prediction.data[compound_uri] = [] unless prediction.data[compound_uri] tuple = { - :classification => classification, - :confidence => conf, - :similarities => similarities, - :features => compound_matches - # uncomment to enable owl-dl serialisation of predictions - # url_for("/lazar#classification") => classification, - # url_for("/lazar#confidence") => conf, - # url_for("/lazar#similarities") => similarities, - # url_for("/lazar#features") => compound_matches + File.join(@@config[:services]["opentox-model"],"lazar#classification") => classification, + File.join(@@config[:services]["opentox-model"],"lazar#confidence") => conf, + File.join(@@config[:services]["opentox-model"],"lazar#similarities") => similarities, + File.join(@@config[:services]["opentox-model"],"lazar#features") => compound_matches } prediction.data[compound_uri] << {feature_uri => tuple} end @@ -168,11 +275,18 @@ post '/:id/?' do # create prediction prediction = OpenTox::Dataset.new prediction.creator = lazar.uri - prediction.title = URI.decode YAML.load(lazar.yaml).dependentVariables.split(/#/).last - prediction.title += " lazar classification" + dependent_variable = YAML.load(lazar.yaml).dependentVariables + prediction.title = URI.decode(dependent_variable.split(/#/).last) + case dependent_variable + when /classification/ + prediction_type = "classification" + when /regression/ + prediction_type = "regression" + end if compound_uri - lazar.classify(compound_uri,prediction) unless lazar.database_activity?(compound_uri,prediction) + # AM: switch here between regression and classification + eval "lazar.#{prediction_type}(compound_uri,prediction) unless lazar.database_activity?(compound_uri,prediction)" LOGGER.debug prediction.to_yaml case request.env['HTTP_ACCEPT'] when /yaml/ @@ -183,12 +297,13 @@ post '/:id/?' do # create prediction halt 404, "Content type #{request.env['HTTP_ACCEPT']} not available." end -elsif dataset_uri + elsif dataset_uri response['Content-Type'] = 'text/uri-list' task_uri = OpenTox::Task.as_task do input_dataset = OpenTox::Dataset.find(dataset_uri) input_dataset.compounds.each do |compound_uri| - lazar.classify(compound_uri,prediction) unless lazar.database_activity?(compound_uri,prediction) + # AM: switch here between regression and classification + eval "lazar.#{prediction_type}(compound_uri,prediction) unless lazar.database_activity?(compound_uri,prediction)" end begin uri = prediction.save.chomp |