Package chemaxon.descriptors

Class MDMetricOptimizer

java.lang.Object
chemaxon.descriptors.MDMetricOptimizer
@PublicApi public class MDMetricOptimizer extends Object

Optimization of metric parameters for dissimilarity based screening. Parameter optimization is uniform for each metric: it is based on the optimization of the performance of the metrics when used for screening molecules. Optimization is performed defining the parameters as variables, its goal is to maximize the selected evaluator function of the given MDHitEvaluator object during screening performed on a test set. The test set consists of actives (similar set) mixed with molecules randomly selected from the target set of screening (dissimilar set).

Parameters are written into the common MDParameters object, determined from the query descriptors.

The MDHitEvaluator object must be set up, for its usage see MDHitEvaluator. Parametrized metrics must be set (weighted, asymmetric, scaled, etc.) MDReaders also must be set up, for usage see MDReader.

Typical usage:

 MDMetricOptimizer metrOpt = new MDMetricOPtimizer( evaluator );
 metrOpt.optimizeParameters( pfp2DIndex, euclideanIndex, similarReader, dissimilarReader );
Since:
JChem 2.0

  • Field Summary

    Fields
    Modifier and Type
    Field
    Description
    float
    maxErrorInF
    Maximal allowed difference in last two calculated values of the enrichment ratio function for termination.
    float
    maxErrorInX
    Maximal allowed dofference in last two calculated values of the weight variables for termination.
    int
    nrOfIterations
    Maximal allowed number of iterations
    float
    percentage
    Consider enrichment ratio values only in cases, when at least the given percent of similars is selected as hits.

  • Constructor Summary

    Constructors
    Constructor
    Description
    MDMetricOptimizer(MDHitEvaluator evaluator)
    Constructor of parameter optimizer.

  • Method Summary

    Modifier and Type
    Method
    Description
    void
    optimizeParameters(int mdIndex, int metricIndex, MDReader similarSet, MDReader dissimilarSet)
    Finds optimal parameter values for the given metric.

    Methods inherited from class java.lang.Object

    clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

  • Field Details

    • nrOfIterations

      public int nrOfIterations
      Maximal allowed number of iterations

    • maxErrorInF

      public float maxErrorInF
      Maximal allowed difference in last two calculated values of the enrichment ratio function for termination.

    • maxErrorInX

      public float maxErrorInX
      Maximal allowed dofference in last two calculated values of the weight variables for termination.

    • percentage

      public float percentage
      Consider enrichment ratio values only in cases, when at least the given percent of similars is selected as hits.

  • Constructor Details

    • MDMetricOptimizer

      public MDMetricOptimizer(MDHitEvaluator evaluator)
      Constructor of parameter optimizer.
      Parameters:
      evaluator - Evaluator of acive/inactive hits

  • Method Details

    • optimizeParameters

      public void optimizeParameters(int mdIndex, int metricIndex, MDReader similarSet, MDReader dissimilarSet) throws MDReaderException
      Finds optimal parameter values for the given metric. Parameters are set and can be withdrawn from the common MDParameters object.
      Parameters:
      mdIndex - Index of molecular descriptor
      metricIndex - Index of metric
      similarSet - Reader of the set containing actives to be selected as hits
      dissimilarSet - Reader of target fingerprints
      Throws:
      MDReaderException - if the reader encounter a read error