Module: OpenTox::Algorithm::Similarity

Includes:
OpenTox::Algorithm
Defined in:
lib/algorithm.rb

Overview

Similarity calculations

Instance Attribute Summary

Attributes included from OpenTox

#metadata, #uri

Class Method Summary collapse

Methods included from OpenTox::Algorithm

gauss, median, #run, #to_rdfxml

Methods included from OpenTox

#add_metadata, all, #delete, #initialize, #load_metadata, login, text_to_html, #to_rdfxml

Class Method Details

.euclidean(properties_a, properties_b, weights = nil) ⇒ Float

Euclidean similarity

Parameters:

  • properties_a (Hash) —

    Properties of first compound

  • properties_b (Hash) —

    Properties of second compound

  • weights (optional, Hash) (defaults to: nil) —

    Weights for all properties

Returns:

  • (Float) —

    (Weighted) euclidean similarity



117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
 
# File 'lib/algorithm.rb', line 117

def self.euclidean(properties_a,properties_b,weights=nil)
  common_properties = properties_a.keys & properties_b.keys
  if common_properties.size > 1
    dist_sum = 0
    common_properties.each do |p|
      if weights
        dist_sum += ( (properties_a[p] - properties_b[p]) * Algorithm.gauss(weights[p]) )**2
      else
        dist_sum += (properties_a[p] - properties_b[p])**2
      end
    end
    1/(1+Math.sqrt(dist_sum))
  else
    0.0
  end
end

.tanimoto(features_a, features_b, weights = nil) ⇒ Float

Tanimoto similarity

Parameters:

  • features_a (Array) —

    Features of first compound

  • features_b (Array) —

    Features of second compound

  • weights (optional, Hash) (defaults to: nil) —

    Weights for all features

Returns:

  • (Float) —

    (Weighted) tanimoto similarity



94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
 
# File 'lib/algorithm.rb', line 94

def self.tanimoto(features_a,features_b,weights=nil)
  common_features = features_a & features_b
  all_features = (features_a + features_b).uniq
  common_p_sum = 0.0
  if common_features.size > 0
    if weights
      common_features.each{|f| common_p_sum += Algorithm.gauss(weights[f])}
      all_p_sum = 0.0
      all_features.each{|f| all_p_sum += Algorithm.gauss(weights[f])}
      common_p_sum/all_p_sum
    else
      common_features.to_f/all_features
    end
  else
    0.0
  end
end