The major goal of the task was to test the prediction power of the QSPR models built on a real-life data set (10,000 chemical structures) under the frame of the previous test phase, model building. The models built on approximately 20,000 compounds have been used in molecular engineering scenarios. The test results on the key part of the molecular engineering process, i.e. designing structures with targeted properties and further improvement of the SPR model for cytotoxicity, and modification in the Custom Data Repository (CDR) is recorded.
Besides testing the molecular engineering process, a comprehensive investigation of the universal structure enumerator, and the property/activity prediction component, as well as the usage of the fragment descriptors and the corresponding descriptor models will also be accomplished. All tests will be made using workflows allowing repeatability of any test cases.
Back to Testing