Predictor
Welcome to the Predictor webserver
- Choose model -- Use the dropdown menus to choose the property you wish to model :
- Select a general kind of property - Here you can select the category of the property you wish to model.
- Select a property to model - Choosing a general kind of property will fill this dropdown menu with the list of available and modellable properties.
- "ColorAtom" option :
- Generate images of the query(ies) with ColorAtom - indicates the influence of each atom in queried molecule(s) using color code.
- There are 2 ways of inputting molecules :
- Draw a single molecule - You can draw a single molecule, or upload a .mol file of a single molecule in the sketcher.
- Upload a file - If you with to use the model on more than one molecule, then you should upload a .sdf file containing your molecules.
- Description of output columns provided at the end of the prediction process :
- Predicted value - predicted class label or numeric value depending on the chosen model.
- Applied models - number of applied individual models / total number of individual models in the consensus model.
- Prediction confidence - prediction confidence label represented in a color code.
- Green - Optimal
- Blue - Good
- Orange - Average
- Red - Unreliable
- 2D structure - 2D structure of compounds after standardization. If "ColorAtom" option was toggled, atoms in the structure will be colored based on their influence on the property according to the model.
- For classification models : a color reflects the contribution to the predicted class. The darker the color the higher the contribution.
- For regression models : a color reflects its positive (blue) or negative (red) increment to the modeled property.
- Comments - detailed information about the prediction.
- AD satisfaction shows the percentage of individual models for which applicability domain was satisfied.
- For classification models, repartition of classes represents the agreement of predicted classes among the applied models.
- For regression models, standard deviation over the predicted values of each applied model is given.