ACD/PhysChem Accuracy Extender (LogP module)  Here and below: Click image to magnify

ACD/PhysChem Accuracy Extender (AE) is a program that calculates new increments for the ACD/LogP and ACD/Solubility algorithm. ACD/PhysChem AE helps advanced users to considerably improve the accuracy of logP or Solubility prediction for the required classes of organic compounds.

ACD/PhysChem AE (LogP module) allows you to compile a LOGPUSER.CLC file that contains new calculated increments for your special class of compounds. Use this file along with the standard LOGP.CLC in order to increase the accuracy of calculation for logP and related properties in ACD/LogP DB, ACD/LogD Suite, ACD/Solubility DB, ACD/LC and GC Simulators, and ACD/PhysChem Batch. The speed of calculation using LOGPUSER.CLC is higher than the previously available system training with user databases.

Benefits of ACD/PhysChem Accuracy Extender (LogP module):

• The possibility to use both experimental logP and logD values as source data. When you use experimental logD values, the program recalculates them into the corresponding logP values. These values are divided into two categories according to a tolerance limit between the experimental and calculated values that you can specify yourself. If you have both experimental logP and logD for your molecules, you can choose any one of them, or use the average.

• Flexible search procedures allow you to specify which molecules to include in the statistical analysis that produces new increments. Search criterion can include information on the types of fragments and chains, the number of known and unknown increments, and chemical structure and substructure queries.

• A set of molecules used for the regression calculation can be changed at any timehowever, the regression should be recalculated when the data is changed.
   
• The calculated increments for fragments and chains can be edited, but you are not able to edit internal increments (i.e., those that are already known to the ACD/LogP algorithm).
   
• The confidence interval for each increment is also calculated. This confidence interval will be further used for the calculation of confidence intervals for the predicted logP and logP-related values.

• You can work with the same project for a long time by adding new data, and deleting or editing already existing data. A snapshot system allows you to save each stage of your work and return to it later; this can be useful when choosing one of the ways to accomplish your task.

• Tautomeric forms can be checked for each structure when added to the database of structures or used for structure or substructure search.
   
• The program is optimized for large databases containing tens of thousands of molecules. The step-regression procedure is also optimized to be quick and require a minimum of system resources.

Here's how the program works:

• To begin, create a new project - this will automatically include the interdependent databases of structures, fragments, and chains. From the very beginning, the database of fragments contains all the structural fragments known to the current version of ACD/LogP software.

• Next, you must fill in the database of structures with your molecules and corresponding experimental data (either logP or logD values). The structures can be imported from an SDfile containing experimental data and/or added from the ChemSketch window. Note that experimental logP or logD values can be added and edited manually in the User Data subwindow of your project.

• Then, during the set-structure-elements procedure, ACD/PhysChem Accuracy Extender (LogP module) will find all of the fragments and chains (interactions between fragments) in your molecules, both known and unknown to the current ACD/LogP DB algorithm. All the fragments and chains will be added to the corresponding databases automatically.

• If you think that your experimental observations are not completely described with the standard and new increments, you can manually define your own user fragments. This can considerably increase the accuracy of logP/logD prediction, for cases where the structural class contains a compound core that ACD/PhysChem AE (LogP module) treats as several small groups or chains (e.g., steroids, sugars, quinones, etc.), or the experimental logP values are influenced by unaccounted intramolecular interactions. You can add new fragments from the ChemSketch window, just draw the molecule and indicate what part of the molecule is to be treated as separate fragment.

In the ACD/LogP Accuracy Extender Preferences dialog box, you should indicate the location of experimental logP or logD and choose what experimental data is to be used.

Create a list of structures for the calculation of increments for new fragments and chains from your molecules.   Start calculating step regression.
After the regression has been calculated, you can update the new increments to your project. The updated increments will be included in the resulting LOGPUSER.CLC.

Note: All of these processes are realized through the User's Wizard, which helps easily use experimental data for system training and utilize this information for prediction of new molecules.