Fast Binary Classifiers for Library Shaping

Automated compound classification is a further approach for rapid library design. 

Molecule encoding. Several molecular attributes, such as predicted and measured properties and structural descriptors, span a high-dimensional feature space. The selection of descriptors is mainly driven by the experience of the scientists, project-specific considerations, and existing knowledge about putative structure-activity relationships. 

Class assignment. If activity values are available at this stage, the tested molecules can be assigned class labels representing activity classes (e.g., low, medium, high). Otherwise, classes can be automatically assigned by analysis of the data distribution, e.g., by cluster analysis. 

Feature extraction. A small set of class-discriminating features is selected (extracted) from the descriptor space, which provide the basis for activity (class) predictions. Traditional feature-extraction methods are based on factor analysis and projection methods [58], 

---