Dear Said,

The number of molecules required depends on the number of physiochemical variable(descriptors) that are used in 3D QSAR calculation.

For example - to test five variables is about 25-30, for 10 variables 45-50 molecules, for 20 variables 60-65 molecules, and for 30 variables 80-85 molecules.

You should computationally validate model by validation techniques, such as like Y-randomization, and cross-validation.

i hope, these points may be helpful for you

Thank you, but if the model is validated for 16 molecules, can be considered rejected ? are there the rols for numer of dadat set for pls analysis?..for qsar 2D it's ok that number of molecules must be more than 30 in order to respect normal low ,but for 3d-qsar Idont know

Dear Said

to get a statistically significant model we need at least 18 compounds

Thank you so much deear Dr Shaheen Begum , but we can validate this mofel for confirming its robustness and power predictiv as well as Y-Randomization test and Tropsha and Goldbreikh validation. Iven if using PLS regression the number of studied molecules muast be more than 18???