SCIGRESS - Computational Tools - QSAR & QSPR Module
SCIGRESS combines experimental data with over 100 descriptors to develop QSAR protocols, and additional descriptors can be defined within the system. As the scientist selects descriptors, SCIGRESS shows their definitions and offers alternative calculation methods in addition to a standard or default procedure. Once SCIGRESS has guided the scientist in choosing properties and calculation methods, the software goes to work in the background, automatically filling in the Chemical Spreadsheet table while freeing up the scientist to do other work. Calculations which are time-consuming may be sent off to computational resources such as Fujitsu's Server for faster processing.
The user then chooses the type of regression analysis to be performed, and the variables for the QSAR regression. Various strategies can be compared and evaluated based on their correlation coefficients, cross-validation scores, and predicted values. From the QSAR/QSPR model results, properties can be predicted for new compounds, including those lacking experimental data or which have never been synthesized. SCIGRESS models can also be calibrated to account for systematic modeling biases and to increase the confidence of predictions.