Changes in version 3.0 (April 2016)

  • New dataset: Updated models by increasing 2,568 new compounds from the ChEMBL 21PubChem BioAssay 1511 and
    from Li et al. paper;
  • New feature: Web app redesigned from scratch ;
  • New feature: Physico-chemical property Predictor (MW, LogP and TPSA)

Changes in version 2.0 (April 2015)

  • New feature: Prediction on the fly;
  • New dataset: Updated models by increasing 1004 new compounds from the ChEMBL 19 (total 5,984 compounds);
  • New feature: Predicted probability maps on the fly;
  • New feature: Multiclass models for Strong BlockerWeak / Moderate BlockerNon-blocker;
  • New feature: Probability forecast (increase confidence of prediction);

Changes in version 1.1 (04/01/2014)

  • New feature: Updated models by increasing 1440 compounds;
  • New feature: Predicted probability maps for evaluated molecules are available in report;
  • By default we now are offering  a simple model associated with the consensus one at the report;
  • Changed the resampling scheme to an external 5 k-fold randomly repeated 10 times; e.g. 10 x 5-fold CV. To reduce the  variance of the CV result, we had repeat/iterate the CV with new random splits;

Version 1.0 (12/12/2013)

  • resampling scheme to three-way data split (80% training set and 20% test set) associated with 5-fold CV randomly repeated 3 times; e.g. 3 x 5-fold CV;