Sometime ago I asked about validation methods for virtual screening protocols. ROC curve and enrichment factors seems to be the most used methods. But my question now (might seem a little bit silly) is: is it mandatory to use any of these methods in a virtual screening protocol? I'm asking this because I read some articles and thesis that didnt use any of those.
Another question I have is: are those methods only used with ligand-based virtual screening or also with structure-based virtual screening?
Here is our take on the basics one should consider in the field:
https://www.collaborativedrug.com/buzz/2014/05/10/recent-changes-in-cdd-vault-calculations/
-Barry
P.S. Also contributed to OS: https://www.collaborativedrug.com/buzz/2014/05/10/open-source-ecfpfcfp-circular-fingerprints-in-cdk/
As you point out, ROC curves and enrichment factors are validation procedures. They give an idea of how successful the screening was, regardless of which type of screening you used (ligand- or structure-based). But they are retrospective, only useful if you know a priori which compounds are active If your screening was not retrospective, but rather prospective then you can't use them. You should focus then on the active compounds you experimentally found and how many compounds you tested in order to find those compounds.
Thank you all!
Now I'm relieved that my protocol is on the right way.
Paul,
Thanks again for your answer. Now I can finally work in peace :)
- Badges
- Science topic
- Similar topics
- Chemistry
- Biochemistry