The searching for agents for Alzheimer's disease treatment via the system of self-consistent models.

Robust quantitative structure-activity relationships (QSARs) for hBACE-1 inhibitors (pIC50) for a large database ( n  = 1706) are established. New statistical criteria of the predictive potential of models are suggested and tested. These criteria are the index of ideality of correlation ( IIC ) and the correlation intensity index ( CII ). The system of self-consistent models is a new approach to validate the predictive potential of QSAR-models. The statistical quality of models obtained using the CORAL software (http://www.insilico.eu/coral) for the validation sets is characterized by the average determination coefficient R 2 v = 0.923, and RMSE = 0.345. Three new promising molecular structures which can become inhibitors hBACE-1 are suggested.