I am curious to learn how PP users employ PCA to create activity prediction of classification models.
For example, take a look at the PP protocols, "01 Learn PCA Model from Asinex," and "02
Use PCA Model and Scatter Plots to View Activity Classes." Aside from the resulting
visualization in step # 02, how does oneactually use the generated model to predict whether
or a molecule not in the trainingset is an estrogen antagonist, prostaglandin, MAO inhibitor, or none of these?
Once the PCA model is created, one can read in molecules not in the training set and obtain
PCA coordinates, but how then to process this information? How does one make a classification
or an activity prediction using the PCA coordinates.
I welcome comments, suggestions, user experiences, references to publications, etc.
Regards,
Jim Metz
