Dear Nail,

sorry for the delay but I missed your question.

Dragon includes only a portion of the existing molecular descriptors but as you stated they are a lot.

We provide a good portion of the molecular descriptors proposed during the years, is up to the user to identify the descriptors that are suitable for its goal.

Additionally an easy understandable descriptor is not necessarily a good descriptor to predict an end-point, otherwise statistical methods should not be necessary, but a good chemist should be enough. Molecular descriptors capability is that they are able to highlight molecular aspects that are not highlighted by a simple chemical structure interpretation.

You should take into account that Dragon can be used to do research, and doing research on molecular descriptors and QSAR means also the analysis of molecular descriptors. Additionally Dragon is used for industrial research, not necessarily the goal is to find the descriptors that are easy to relate to a physical chemical property or easily interpretable. Sometimes the best statistical model is the goal and the interpretation is an add-on.

I always suggest not to try to find a general meaning of a complex descriptor but it is more useful to find the meaning for a complex descriptor in the dataset that is used to build the model. Analysing the descriptor behaviour in a defined dataset can help the user to understand its chemical meaning.

Dear Nail,

sorry for the delay but I missed your question.

Dragon includes only a portion of the existing molecular descriptors but as you stated they are a lot.

We provide a good portion of the molecular descriptors proposed during the years, is up to the user to identify the descriptors that are suitable for its goal.

Additionally an easy understandable descriptor is not necessarily a good descriptor to predict an end-point, otherwise statistical methods should not be necessary, but a good chemist should be enough. Molecular descriptors capability is that they are able to highlight molecular aspects that are not highlighted by a simple chemical structure interpretation.

You should take into account that Dragon can be used to do research, and doing research on molecular descriptors and QSAR means also the analysis of molecular descriptors. Additionally Dragon is used for industrial research, not necessarily the goal is to find the descriptors that are easy to relate to a physical chemical property or easily interpretable. Sometimes the best statistical model is the goal and the interpretation is an add-on.

I always suggest not to try to find a general meaning of a complex descriptor but it is more useful to find the meaning for a complex descriptor in the dataset that is used to build the model. Analysing the descriptor behaviour in a defined dataset can help the user to understand its chemical meaning.