Predictor plugin example?

Does anyone have an example using the calculator plugin "Predictor"? At least an example with data of going in? The description within Knime is a bit short and I can't seem to find other sources. Thanks.

Please refer to the following sites.

http://www.chemaxon.com/marvin/help/calculations/predictor.html />http://www.chemaxon.com/marvin/help/calculations/predictor_training.html

Best,
Taka

So I need to use cxtrain to create my trainingsfile? I can't just use my own sd file to start with?

Also, to point out, the text in the help page seems to missing something:

 

2. Execute the following command from command line:

cxtrain prediction -t MP -i meltingpoint my_data_mp.sdf

The data tagged MP is    ...

 

.... is what?

 

In general, it would be nice with some visual examples of nodes. Not all users are super users who grasp all details at once, even when digging (and I use "digging" in double meaning) through the available manuals.

Hi,

I'm sorry for the inconvenience I've caused you.

Yes, you have to create your training file first.
In general, you can find the cxtrain command line tool in the following directory.

C:\Program Files\ChemAxon\JChem\bin

For instance, if your training data (SDF) have property as logP named, you can execute the following.

C:\Program Files\ChemAxon\JChem\bin>cxtrain.bat prediction -t logP -i logP_model sample_data.sdf

-t:property name which you would like to use for prediction.
-i:training id which you can see in the list of Predictor node.

Then, you can see the created training file in the following directory.

<HOME_DIR>\chemaxon\calculations\training

Best,
Taka

Thank you very much. I will try to work with that example.

Some observations while working with this:

a) the trainind_data file has to be in the original folder where cxtrain puts it. To have more flexibility,  it would be nice to see a dialogbox to select the training file from where ever it might be. it could still use the current path as a default path if nothing else is choosen?

b) on mac-os, the path to the training file btw is:

/users/USER/.chemaxon/calculations/training (note that the chemaxon is a  hidden folder)

c) one can not use the system default for decimal points, cxtrain seems to only accept ".", not "," else one receives an error. I am not sure at this point how the remining chemaxon software handles the decimal point, but this is quite an inconvenience.

 

Dear Rhenium,

Sorry for the late answer. The Calculator Team is quite busy nowadays.

The answers for your questions :

i) We currently do not plan any feature developments concerning the Predictor plugin. I admit the lack of flexibility the  plugin has, but being a minor issue now we can't deal with that now.

ii)  Yes, you are right. On each UNIX-based system (Linux and MacOS as well) the path to the generated training file is /USER/.chemaxon/calculations/training.

iii) Regarding the input ChemAxon tools accept, I could say that only decimal values with the '.' separator are valid.  Regarding the output, I could say that the generated separator is usually based on the system default value. However, the globally accepted notation in the scientific literature is the '.' separator.  If you want your input to be recognized uniformly by ChemAxon tools, please use the '.' .

I hope I answered all your questions.

Daniel