The attached protocol shows how a multi-category Bayesian model can be derived from ChEMBL data as I already alluded to in a previous post (https://community.accelrys.com/message/7187#7187) but which I never got around to actually post. It builds a model for each human target for which at least 10 actives are available that pass some loose structural and physical property filters.

It is assumed you have a local copy of ChEMBL. On my laptop I have installed the Oracle version of ChEMBL version 7, version 8 should work as well but has not been tested. You most likely have to change the component in pipeline 2 so that it points to the right data source.

Cheers,

Willem