Frameworks
This new method, intended in particular for large data sets, generates classes with scaffolds based on ring systems and connecting chains, similar to the so-called “Murcko assemblies” described by Bemis and Murcko (Bemis 1996). A major distinguishing feature of this method is that each molecule yields just one scaffold, so it is guaranteed that each molecule will be contained in only one class. Furthermore, the method requires very little machine memory and is extremely fast, so is appropriate for large data sets.
A principal advantage of the frameworks method is that each class is generally very homogeneous, since each molecule is represented by its major ring system. However, this benefit often has the downside of increasing the total number of generated classes relative to other methods like Ring Systems (see below). In an effort to address this problem, the classes generated using the frameworks method are automatically organized such that similar scaffolds appear next to one another in the spreadsheet. This makes it much easier to ascertain the overall structural diversity present in a data set. Note that this post-processing step is unique to the frameworks method of class generation; for other methods, you will need to use the Generate Class Families feature to organize scaffolds by structure.
Be aware that the single-scaffold-per-molecule characteristic of the frameworks method means that it does not utilize the Homogeneity or Redundancy settings discussed in the section below.
In order to help reduce the number of generated classes, a new Consolidate Classes option was introduced in ADMET Predictor version 9.5. This option is available on the Class generation options dialog, which is displayed by clicking the Change General Options button on the Generate classes dialog. When this option is selected, a post-processing operation is performed on the initial set of classes in an effort to reduce the total number. Each class (referred to here as the "donor" class) is evaluated using every other class having a smaller scaffold (referred to here as "receiver" classes). If a receiver-class scaffold matches all compounds in the donor class, the donor-class compounds are added to the receiver class, and the donor class is eliminated. Each donor class can be added to only one receiver class, so when the process is complete it is still guaranteed that each compound belongs to only a single class. The Consolidate Classes option can significantly increase the time required for class generation, so it should only be used with data sets containing fewer than roughly 100,000 compounds.