Training of the LogD plugin

If you think your experimental data could improve the performance of the default logD calculator, you can take advantage of the supervised logD learning method that is built into the calculator.

LogD calculations based on trained pK_a and logP data

The logD method can be trained by applying existing pK_a and logP training sets. For detailed information on training the pK_a and logP plugins, see the corresponding pK_a and logP training manuals. To demonstrate the training, tests have been run with the same datasets used for the pK_a and logP calculations.

Application

Marvin

1. Choose MarvinSketch menu : Calculations > Partitioning > logD.
2. Select the User defined or the Weighted method to activate the training option.
3. If you have many logP training sets, you can select the one you want to use for training from the logP training ID dropdown list.
4. If you have many pK_a correction libraries, you can select the one that you want to use for training by selecting the 'Use the pK_a correction library' option, and choosing the library from the dropdown list.




Test results

Trained value :

pH	logD

7.40	-0.34

Untrained value :

pH	logD

7.40	-0.08

cxcalc

To apply your pK_a correction library to train the LogD method, use the --pkacorrectionlibrary option :

cxcalc logd --method [method] --pkacorrectionlibrary [library name] [input file/string]

To apply your LogP dataset to train the LogD method, use the --method main option, combined with the --logptrainingid secondary option :

cxcalc logd --method [method] --logptrainingid [library name] [input file/string]

Example

1. $ cxcalc logd --method user --pkacorrectionlibrary mypka_1 --logptrainingid mylogp_1 --pH 7.4 "CC1=NC2=C(N1)C(O)=NC(N)=N2" (trained)

2. $ cxcalc logd  --pH 7.4 "CC1=NC2=C(N1)C(O)=NC(N)=N2" (untrained)

Test results

Trained value :

id	logD[pH=7.4]

1	-0.34

Untrained value :

id logD[pH=7.4]

1	-0.08

Chemical Terms

Evaluator

The parameters pkacorrectionlibrary and logptrainingid are utilized in Chemical Terms Evaluator as well :

 evaluate -e "logd('method:[method] pkacorrectionlibrary:[library name] logptrainingid:[id]')" [input file/string]

Example

1. $ evaluate -e "logd('method:user pkacorrectionlibrary:mypkalib_1 logptrainingid:mylogp_1 pH:7.4')" "CC1=NC2=C(N1)C(O)=NC(N)=N2" (trained)

2. $ evaluate -e "logd('pH:7.4')" "CC1=NC2=C(N1)C(O)=NC(N)=N2" (untrained)

Test results

Trained value :

7.4;-0.34

Untrained value :

7.4;-0.08

Instant JChem

You can also apply your pK_a and logP training libraries via Chemical Terms in Instant JChem .
1. Choose the 'New Chemical Terms Field' icon on the panel on the right side.
2. Type the chemical term into the window, use the method, pkacorrectionlibrary and logptrainingid parameters. Do not forget to specify the Name, the Type and the DB Column Name parameters.

Example
The following snapshot shows the usage of the LogD training in the 'New Chemical terms' window. The expression

logd('method:user pkacorrectionlibrary:mypkalib logptrainingid:mylogp', '7.4')

defines that the plugin use the mypkalib pK_a correction library, the mylogp logP training library and pH value 7.4.



Part of the result table is presented below. You can see the difference between the untrained and the trained logD values.