DOCK Blaster:Reliability

From DISI
Revision as of 18:22, 21 September 2007 by JohnIrwin (talk | contribs)
Jump to navigation Jump to search

Should I take the results produced by DOCK Blaster seriously?

Whether you are an experienced docker or completely new at this, you may be skeptical about DOCK Blaster and docking in general. After all, docking has a famously high false positive rate for both sampling and scoring reasons.

Should I "trust" a docking hit list from DOCK Blaster?

We have designed DOCK Blaster to perform calibration calculations before docking. Depending on the information you supply, various controls and tests are performed and reported on. Whereas every case will be different, here is a rough guide to interpreting the results of calibration.

there will always be false positives, probably many

Your mission is to select compounds from near the top of the docking hit list to purchase and test. Use common sense, buy a range of chemotypes, if possible. watch for solubility. test at high concentration. control for aggregation.

some true positives will probably score poorly

Our automated, high throughput approach to docking and database preparation inevitably lead to some perfectly good compounds scoring poorly for artifactual reasons. Compensate by looking at how similar compounds scored.

you can obtain helpful diagnostics with a docked control ligand in mol2 format

you can obtain even more diagnostics if you have actives and inactives (controls)

without a docked ligand or annotated ligands, evaluating docking performance is challenging

References

  • the DUD paper, Huang, Shoichet & Irwin J Med Chem 2006



pasted in


You are wise to be cautious about trusting the results produced by DOCK Blaster. Molecular docking and virtual screening methods are notoriously susceptible to many possible failures due to the numerous approximations made in the calculations. We recognize this inherent skepticism, and have therefore attempted to provide tools with which you can reassure yourself that DOCK Blaster results really are worth investigating - or not, as the case may be.

Here is the evidence we will gather to investigate the calculation:

  • 1. score of the supplier ligand in its supplied pose
  • 2. ability to re-dock the supplied ligand
  • 3. score & pose of similar ligands, drawn from ZINC
  • 4. score & pose of property-matched, dissimilar ligands drawn from ZINC
  • 5. score of extrema ligands
  • 6. score of actives, if supplied
  • 7. score of inactives, if supplied
  • 8. appearance of top scoring ligands from the lead-like or fragment-like database.

Now, we review how to evaluate the above results in order to estimate whether the results are worth pursuing, for example, by purchasing compounds.