Methods and Descriptors

Our Methods and Descriptors:

Data selection and neural net based QSAR modeling are the key methods in constructing robust predictors with large datasets. With the exception of the blood-brain barrier partition dataset, all ChemSilico predictors were developed around neural net QSAR models. In the case of CSLogD and CSLogWS, results from several QSAR models were used in expressions to compute LogD and LogWS.

Molecular descriptor space employed a total of 519 topological and E-state descriptors. Of the 519 descriptors, 158 were new proprietary molecular descriptors, developed by ChemSilico for exclusive use in our predictors. Although the utility for the majority of descriptors used in our models has been demonstrated in many important QSAR models published over the last eight (8) years, the proprietary descriptors assisted immeasurably in achieving high quality correlation throughout all our QSARs.

For specific information on ChemSilico methods and descriptors,

please visit the appropriate page by clicking on the button below.

Information on ChemSilico Methods

Information on ChemSilico Descriptors

Methods and Descriptors Reference Page





search
links
user login
contact us