-
Create the SD file:
- Compounds can consist of fragments with no counter-ions. If there are multiple
fragments, GTD uses the largest.
- Provide a value for the response property you plan to model for each compound.
- For a classification model (bayesian, random forest, or pharmacophore), the response
property must be categorical. When you build the model, specify a “preferred” value of
the property, which corresponds to a positive category. For example, for an activity
model, the categories might be “active” and “inactive,” with “active” being the
positive case. For a toxicity model, the categories might be “toxic” and “nontoxic,”
with “toxic” being the positive case.
- For a regression model (random forest regression), the response property must be
numeric. For XC50-type data, the preferred form of the response is pXC50 (the negative
base-10 logarithm of the XC50 value) so that greater values correspond to increased
potency. For example, for an activity model, the name of the response property might
be pIC50.
-
Copy the SD file to your 3DDrive.
|