The answer is "it depends". I have had some success with blind docking, where the entire protein is enclosed within the simulation cell. However, the success rate depends on the types of ligand-protein interactions, the degrees of freedom of the ligand, and how large the protein is. You can get some idea of how well your predictions might be by starting with redocking virtual experiments, where you see how well you can reproduce a known x-ray crystal structure of a ligand similar to yours in complex with a protein receptor.

Thanks Mr. Rudy J Richardson Sir for your answer. I agree with you but here I am trying to predict the target protein, not just a particular site in a specific protein.

Dear Javid,

You can go with inverse docking in order to predict the target but the target must posses the defined binding site. What u are supposed to do is that first re-dock the ligand of the target and based on lowest RMSD (should be less than 2A otherwise reject the target) note the score and then dock the unknown ligand, find the same. Do this for all the putative targets and finally based on lowest rmsd and highest docking score rank the targets. It is better if u will go for the simulation in order to refine your results and justify your target.

Regards

My apologies for misunderstanding your question.

To predict a target protein, there are at least two approaches: inverse docking and pharmacophore mapping. You can find recent papers on these approaches by doing a literature search. In addition, there are web servers for doing pharmacophore mapping, e.g., PharmMapper.

Dear Mohsin thanks for your answer. I am aware of the approach, however I am doubtful about the scoring function particularly with Autodock! I have seen many times higher score is predicted for an inactive compound than a well known drug!

Hi

"wanted to predict the target site"

Your means that active site ?  However I would recommend you use Sybyl-X and "Protomol" to find active site of your target. Its very powerful tools in this field.

The Best

Mohamd

Molecular modeling is not reliable unless confirmed experimentally. No "scoring function" can ensure reliable prediction. I would not volunteer to act as a guinea pig in clinical trials if the drug candidate emerged from molecular modeling alone. And if one has the experiment what use is the molecular modeling then? Nice pictures? Cool animations?

Mr Martin Klvana you are right. However here we are interested in predicting the target protein/site for a new molecule which has been found to be active through in vitro experiments. We were interested in using molecular docking for that. So what would be your suggestion in that case!

there are research papers which describe different protocols for target fishing. These protocols combine structure similarity search with docking and pharmacophore mapping. these protocols are more likely to help with target prediction, yet you must validate these possible targets experimentally.

I do not have much experience with docking and the little I have is probably outdated but there is one potential problem with inverse docking: The absolute scores of the inverse docking of one ligand to different proteins might not be comparable if the binding sites differ substantially. So, for each target one should probably also consider docking of a "reference ligand" for which some experimental binding data are known. Then choose the targets based on the target-specific relative scores . . .

Thanks Mr Martin, that seems pretty logical! will look into that possibility...