We have found few proteins in the cell (not human) and a drug to treat the cell. Is there an insilico way to find on which protein does my drug interacts (eg. Cytoscape)?
This is quite an in depth and huge undertaking. But a worthy one, and one that I am currently working on myself. You will need a 3D structure of these few proteins; try homology modelling first e.g., PSI-PRED->BioSerf, and then if that fails consider something a little bit more heavy-weight such as I-TASSER.
Once you have a structure, and a prepared structure of a ligand (consider using QM calculations to define the 3D structure of your ligand, use something like Gaussian or Ligand Prepare if you have Schrodinger software suite), you can then try to locate active sites with tools such as "Pocket Finder". This will help give you an idea of where drugs are going to try binding (but not which particular drug). Then, with this information, try something like AutoDock or Glide (part of Schrodinger) to see whether a drug fits into any of the proposed active sites.
There is a huge amount to learn in those few steps. Good luck :-)
Get all atomic-resolution structures for all proteins produced by the cell: crystal structures (if available), or generate the structures using homology or de novo modeling. Generate all possible conformations of the protein structures using molecular dynamics and/or Monte Carlo simulations. Dock the ligand into each conformation of every protein . . . An impossible task?
While you gather all the necessary (likely to be incomplete) information, begin an empirical experiment to significantly increase your odds to succeed.
@Anthony and @Martin - Thanks for the suggestions.
But generally speaking, the docking is a general concept which means any drug can interact with any protein with higher or lower binding energy but that cannot be exact in living cell. Each drug has its own protein of target. For eg. I have 5 proteins A,B,C,D & E in the cell with binding energies of -2.3, -3.3, -1.2, -4.4 and -3.8 with my drug of interest.
So what would be the appropriate target for my drug?
I am still confused.
@Tausif Alam - So, what could be the better way to understand this through insilico approach?
There isn't, as far as computational methods and their ability to figure our the potential possibilities are concerned.
In silico methology is not magic, you can only get out of it what you (or others) have put into it in the form of rules based on existing information. The advantage is that you can sort through a very large numbers of permutations and combinations.
if your protein structure model is based on relatively fixed (stable) structure and does not include all freedom of movements, you already know that result will be less than accurate. But perhaps by now, the routines used in such analyses have advanced to include many such factors... I'm just not sure!
"...that cannot be exact in living cell. Each drug has its own protein of target. "
What makes you think that?
Thirumal, you can use one of the freely available web services (or all of them) dedicated to the prediction of protein targets for chemical compounds:
http://www.way2drug.com/passtargets/
http://sea.bkslab.org/
http://www.swisstargetprediction.ch/
http://potentia.cbs.dtu.dk/ChemProt/
Also, it may be useful to search for the known bioactivities of your compound (or compound that is similar to yours ) in the databases:
https://www.ebi.ac.uk/chembl/
https://pubchem.ncbi.nlm.nih.gov/
Hope this helps!
@Tausif Alam - That's the thing the lock and key mechanism states.
Hi Thirumal, the "lock and key" mechanism is quite a rigid philosophy. Consider looking into allosteric and induced fitting/binding.
Hi Pavel V Pogodin
The links are quite interesting. Thanks...
But I feel those servers target human proteins. I am working with the virus. Can you help me with that?
"That's the thing the lock and key mechanism states."
Macromolecules couldn't care less what your contrived simplistic mechanism states.
Unfortunately, I'm not aware about ready-to-use tools for the prediction of interactions with virus proteins, which does not mean they are absent, though.
Anyway you always can build your own predictive model(s) which will fit your purpose. If this is an option, I suggest the following:
1) Check out some papers dedicated to the QSAR and Chemoinformatics to get the view of current methods.
2) Find training data - known ligands (generally speaking) for your proteins; - ChEMBL is an option.
3) Use some tools to calculate chemical descriptos for compounds (Dragon, ISIDA, etc).
4) Use some tools to built predictive model (I would recomend to use packages for R-language, they are numerous; or Python libraries). Do not forget to validate the model, it is essential.
6) Select the most probable targets according to the prediction results.
This is a very superficial plan.
Good luck!
@Tausif Alam - yeah.. exactly. It is not so simplistic. But what could be the way to make it comparatively simplistic using in silico approach for my question?
- Badges
- Science topic
- Similar topics
- Biological Science
- Bioscience
- Biotechnology