Hi
In order to get a good QSAR model the number of
compounds in the dataset should be > 20 and also there should be considerable
difference in the range of activities. Ideally the acceptable r2 value is >
0.80 but again it depends on lot of other factors. If the main objective of your
QSAR analysis is make perfect predictions then you should go in for a model with
r2 >0.9, but if your main objective is just to cluster the molecules in
activity ranges then r2 > 0.8 can be used. It is good to have structurally
diversity while building the QSAR model.
Cheers
Amit
----------------------------------------------------------------- Amit Kulkarni, Ph.D. Applications Scientist Life Science Molecular Simulations Inc. Burlington, MA 01803 (781) 359 1217 Fax: (781) 229 989 Email: amit (+ at +) msi.com -----------------------------------------------------------------
|