State Of The Art Of Prediction Of Protein Functionality After Ligand Binding

Question

[Flow]

Current Virtual Screening methods (Docking, Pharmacophore, etc) can screen a large database of chemical compounds and rank them according to some computed score value. Usually one computationally screens a database containing millions of ligands against a given target, and selects a small fraction of the top scoring molecules, so they can be tested experimentally.

Sometimes such methods work, and some ligands bind stronger or weaker than predicted to the protein target. But it most cases, the functional consequence of ligand binding to the protein is not or can not currently be predicted; for instance, for a given ligand some rearrangement could happen in the other part of the protein, which would affect enormously to its biological function, while for some other ligand, the measure affinity could be of the same order of magnitude, but the functional consequences could be completely different.

So my questions here is; do you think in such scenario (ligand database virtual screening), and although affinity could be more or less estimated, can protein functionality (after ligand binding) predicted? And, which are the best tools/related-publications nowadays?

Answer 1

[miquelduranfrigola]

If I understood well, you want to do small-molecule virtual screening for functionality, not just binding. I'll give you a general and two specific approaches.

General Go to a database that has functional bioassays of small molecules (PubChem, ChEMBL...). Select your favorite functional assay. Collect the molecules that have been screened. Do chemogenomics, that is, learn ligand-based rules to do profiling (have a look at Brian Shoichet's Nat Biotech 2008 for an approach, although they use it to predict binding). Note that, for this, you do not need to know the binding mode, not even the target structure, etc. All you do is learning molecular rules to regard chemicals as active (functional) or not.

If you know the mode of action Of course, in case you know the mode of action of a collection of molecules, and ideally the binding site, you can infer that molecules binding to the same binding pocket (allosteric site, or whatever) will produce a similar functional effect. Then, you can do docking, for instance.

If you have all the time in the world You can try molecular dynamics, to explore changes of conformation upon binding, etc, but for this you need to know the system very well.

Definetely, what you want to do is a chemogenomics. The main drawback is that preexisting bioactivity data is required.

[Flow]

Thanks and very interesting. From an ab initio point of view, I would then like to use Molecular Dynamics in the way you mention. Could you point to some interesting papers?