Competitive Prediction Accuracy

Due to its industry-unique molecule-space mining methods, ActiPred/Chem delivers competitive performance while at the same time providing structural feedback about the possible reason for activity/inactivity. The following table lists the accuracy for some of ActiPred's built-in models.

Insightful Substructure Highlighting for BBB Permeables

As an example for the additional context information provided by ActiPred, the following table shows some of the molecules from the MDDR dataset that were classified as permeable.

For each molecule, the substructure that contributed most likely to the permeability property is highlighted in red. Moreover, all molecules are sorted from most likely active to most likely inactive.

When scrolling further down in the result list, likely non-permeable molecules will be encountered as shown in this snapshot from the end of the result list:

The molecules on this end have slightly different structures. The highlighting for inactives uses a different color (here blue) for easier visual differentiation.

Insightful Substructure Highlighting for hERG Blockers

As another example for the interesting insights provided by ActiPred/Chem, the following table shows some of the molecules from the MDDR dataset that were classified as hERG blockers.

For each molecule, the substructure that contributed most likely to the hERG blocking property is highlighted in red. Moreover, all molecules are sorted from most likely active to most likely inactive.

The end of the result list will again contain molecules classified with high likelihood as being non-blockers. Fragments contributing to the non-blocking property are again highlighted in blue.

In summary, being able to provide the user with the significant substructures that lead to a specific classification is a crucial differentiator over other tools and can be of great help in lead generation.