Question:

I am using the Jchem tool for excell to predict pka and logD for fluorinated compounds. Generally a lot of QSAR models have very little of these compounds in their training data and therefore the chemicals are often out of domain. I was wondering if there is any in-built domain assessment for pka (acidic and basic) and for logD models. Or if I could find out how many fluorinated compounds (and the extent of fluorination) there are in the training datasets for these models. I can't seem to find anything about the domain in the documentation.

Answer:


There are none. We don't have fluorinated compounds in our training set, mainly because the main domain of our pKa and logP/D predictors (small and middle-size drug molecules) typically doesn't contain such molecules. Fluorinated compounds are usually not measured for pKa or logP/D.


In general our pKa predictor uses an ab-initio model for pKa prediction and doesn't have a training set. Our logP model, however, uses training sets. You can read more about them here:


https://docs.chemaxon.com/display/docs/logP+Plugin