%dl#lazar_algorithm
  %dt
    %code lazar
    prediction
    (
    = hide_link "#lazar_algorithm"
    )
  %dd 
    %p
      %code lazar
      searches the training dataset for  
      = toggle_link "#similarity", "similar"
      compounds 
      %em (neighbors)
      and calculates the prediction from their measured activities.
      %code lazar
      calculates predictions using
      %ul
        %li
          a majority vote (weighted by compound similarity) for
          %em classification
          (
          %a{:href => "http://www.in-silico.de/articles/modi020905.pdf"} original publication
          )
        %li
          a local QSAR model based on neighbors for
          %em regression
          (
          %a{:href => "http://www.in-silico.de/articles/mh_tf.pdf"} original publication
          )

    %p
      = toggle_link "#significant_fragments", "Significant fragments"
      are highlighted in the structure display as follows:
      %ul
        %li
          .active activating fragments
        %li
          .inactive deactivating fragments 
        %li
          .inconclusive regions, where activating and deactivating fragments overlap
        %li
          .other inert parts

    %p
      Please keep in mind that predictions are based on the measured activities of neighbors. 
      = toggle_link "#significant_fragments", "Significant fragments"
      are solely used to determine 
      = toggle_link "#similarity", "activity specific similarities"
      of neighbors.