Overview of allosteric models used in the case studies.

In the case of Protein Kinase B a three-class model was created (natural ligand mimicking peptides formed the third class). For class errors, sensitivity is recall of allosteric small molecules, specificity is recall of non-allosteric molecules, and in the case of protein kinase B a third class error (allosteric biological recall) is added. Likewise positive predictive value (PPV) quantifies precision for allosteric small molecules, negative predictive value (NPV) quantifies precision for non-allosteric molecules, and a third value quantifies precision for allosteric biological in the case of Protein Kinase B. In addition to these parameters, the MCC is calculated. Note that the values for the non-MCC parameters in the three-class model have been scaled to the same range as the binary classification models to allow direct comparison. Abbreviations: MCC – Matthews Correlation Coefficient, HIV – Human Immunodeficiency Virus.