If you need to select single protein-ligand complex for structure based pharmacophore modeling, then you have to concentrate on the ligand present in the active site. The ligand should have the followings..

1. Good binding affinity towards receptor.

2. Capable to modulate the function of the protein (Interacting with potential residues)

3. It can be a drug molecule or lead molecule.

For protein selection,

1.The protein molecule (crystal structure) should be the more reliable one (good resolution structure). The structure should not have any missing residues or mutation in and around the active site.

Selection based on these criteria can be helpful for successful structure based pharmacophore generation of a protein-ligand complex.

Dear Dr., l would to inform you the binding P-L complex modeling depend the disease, .... But in vitro used BSA ( Bovin Serum Albumin ). The binding should be reversible mostly weak bonds.(H-B)

I recommend using more than just one protein-ligand complex for the purpose of pharmacophore models generation. Pharmacophore modeling is not computationally demanding so that you can generate poly pharmacophores in a relatively short time. Doing so, you can thoroughly explore the screened libraries for potential chemically diverse modulators.

Select complexes where your protein of interest is co-crystallized with ligands representing different chemical classes, and are experimentally tested and classified as being potent modulators of the protein's activity. Take into account also to select ligands that interact with different essential binding features within the binding site. The chosen complexes should have high resolution and have their binding sites well defined without any missing residues within the binding site.

Having the pharmacophores generated and the libraries virtually screened, filter your screening results according to the required criteria, but ensure to remove duplicates.

If you wish to perform molecular docking afterward, choose the best quality protein structure whose resolution is the highest and whose number of missing residues is the least.

@Sondos. Suppose you have been given 15 complexes. Only 5 protein-ligand complexes have binding affinity in nM (Ki) with moderate resolution. The next 5 complexes have good resolution (less than 1 A) and do not have binding affinity report. The final 5 complexes contain drug-like/lead molecules with no affinity and moderate resolution. You need to select only one protein-ligand complex. Which category will you choose and why?

By my opinion I would select first one, because we focused on molecular recognition of a biological target shared by a group of compounds(active molecules with good binding affinity). The second group can be considered to validate your result (you can dock the same compounds here).

Prasanth Kumar

I agree with Lakshmanan Loganathan in selecting a complex from group 1. What you are actually concerned with when generating a pharmacophore model is the chemical features of the ligand and the receptor-ligand interactions that led to potent inhibition/activation. The resolution is not a big issue here.

The only purpose of generating pharmacophore models is to use them as search queries that reduce the search space and aid in selecting compounds having the ability to bind the receptor with high affinity. Virtual screening results would be filtered later, so you can exclude compounds that are not drug-like.

You can, however, select a protein structure other than the one used for pharmacophore generation, and utilize it for docking, should you be looking for better resolution.

Either way, a resolution better than 2.5 Å is acceptable for structure-based drug design, so you shouldn't restrict yourself to be looking for structures whose resolution is better than 1 Å.

I hope this is helpful.

Hi prasanth

This is a very interesting question as many researcher wisely choose wrong model/complex for computational studies. As I am a computational and crystallographer i think i may in position here to gv u a little clear idea.

The criteria for selecting a single complex from the case you suggessted may be as mentioned here

1) if all the 5 PL complexes have same binding pattern, than u can choose any of them based on good resolution

2)secondly in pdb one have to often see which are the cocrystallised ligand,ions and cofactor. Also, is the pdb completely mapped to complete length of protein i.e. is electron density of few residues are missing or few aa not modeled.(if this is case avoid this pdb)

3)Ki, Km or any binding aff measure are very relative values. Actually one have to read paper that how this values were calculated? E.g. i hav seen a case where Kd is 1uM through biophysical assyas but the drug was not able to penetrate cellwall in invivo. Now if choose this basis for pharmacophore it will be completely incorrect from biological point of view

There are many ways one can select pdb let ne know if you want to have more discussion on this. I am nt a real expert of pharmacophore though.

Dhaval Patel . This is an interesting point of view from a crystallographer. But, how would you decipher that the chemically different ligands will have the same binding pattern. There may be consensus in intramolecular interactions and subtle changes in the interactions are widely used as variables for receptor-based QSAR modeling. Here, we are expecting that there will be some subtle changes which can describe variance among the PL complexes in modeling the binding affinity through QSAR.

I strongly agree with the avoidance of PDB structures containing missing residues/lack of electron density data upon the amino acid residues/ligand present in the protein-ligand interface. According to point 2, this task may need considerable amount of time as one has to go through all the PL complexes for a particular target of interest in the PDB. There are programs available which can readily detect missing residues and mapping of electron density data. Given that structure-based pharmacophore is an abstraction process (we can say low-resolution query in lieu of @Sondos Musleh), structure preparation in the initial stage will alleviate this problem although partially.

As per discussions here, we can safely eliminate resolution as a factor provided that it should be reasonable (

To the best of my knowledge, cellular activity is not a factor to be taken into account when choosing a ligand to work with. Even if you construct your pharmacophore based on a cellularly active ligand, this doesn't guarantee that the ligands obtained from virtual screening would also be cellularly active. The cellular activity is a pharmacokinetic property that is not directly related to the functional groups the ligand has. For instance, suppose your pharmacophore, which was built from a potent intracellularly active ligand, has 5 features, a hydrogen bond donor, acceptor, ring aromatic, and 2 hydrophobic features. The virtual screening retained two hits, both contain all of the required features, with fit values above 4.5, but the first has an ionizable group and hence is a polar compound, and the second has long carbon chain and hence very lipophilic. The two compounds will have different pharmacokinetic profiles, and hence different intracellular activities, even though both would be active in the biochemical assay.

In every virtual screening occasion, you will end up with many ligands with a wide range of activities, both in the biochemical and cellular assays. Should your most active ligands be inactive intracellularly, then you can pursue an optimization procedure, where you retain all functional groups required for activity, but make other changes to improve the pharmacokinetic profile.

@Sondos Musleh, I strongly agree with your point that pharmacokinetic properties could not be directly associated with pharmacophore. Hence, it is not a factor to be concerned with.

Very nice discussion.

As per my understanding, crystal structure with reasonable resolution and binding affinity that have slightly consensus profile in terms of binding residues (if possible) should be the top criteria. I am stressing for the binding residues (not the ligand functional group) because biochemical and cellular activity is encoded by the interactions (allosteric/competitive or agonist/antagonist) from such domains/pocket. Remaining things mentioned here can be secondary.

I would be happy to read more discussion for this query.