Hello,
I am currently working on a project with shared models. Essentially we are training models on proprietary data sets and sharing the models, but not the data sets. Domain applicability can then become a little tricky.
I am trying to estimate fingerprint coverage for test set compounds. I can do this for models which I have generated by using the Learned Model Reader component, exporting Bin Information, and calculating the coverage for each molecule in a test set.
Unfortunately this only works for models I have built as the Learned Model Reader seems to require a learned model and not the exported component we are sharing. Is there a way for me to get to the bin fingerprints for the exported model component?
From some reading I gather that Accelrys has done a lot of nice work with domain applicability and this can be included in the original models with options such as Track Property Ranges and Perform OPS Analysis, however the original models were shared without these options.
Many thanks,
Andreas