The pharmacophore is generated from structure and ligand-based drug design. How can we know the pharmacophore generated is valid for virtual screening? What the method can used?
It is necessary to validate your Pharmacophore before employing into virtual screening. There are numerous methods and metrics for validation of both structure-based and ligand-structure based pharmacophoric models.
if you are using commercial software like PHASE in Schrodinger suite or drug discovery studio, they have inbuilt criteria for validating your pharmacophore models.
However, screening with decoy set & ROC will validate your pharnmmacophore model.
Personally I am not convinced with ROC validation because it is valid to perform this experiment when In-actives and actives are known by enzymatic essay (not cell based assay).
No matter how you generated your pharmacophore model, you should always validate it. The same holds for your virtual screening method.
Make a test database containing real actives and a much larger number of decoys. Chose those decoys so that they have similar "generic" properties when compared to your actives (e.g. similar Mw, number of donors and acceptors,...). Then screen this database of actives + inactives and analyze your results. Does your model successfully retrieve the actives? Do you get some decent enrichment form your selection? Are there specific classes of compounds that are better retrieved? (or not?).
Notes:
1) Indeed, if you have compounds that you know to be inactive (because they have been tested against your target), it is obviously a good thing that you include those in your validation set.
2) Ideally, the 3D structures you used for generating your pharmacophore should not be used as such for your validation. Even those compounds should be brought back to a 2D file, as if it was an unknown molecule. Then, you must go over the full pre-treatment and conformer generation process for those compounds as well.
- Badges
- Science topic
- Similar topics
- Filing