Hey everyone, I was just wondering if anyone knows of any research that has been conducted to determine a way to theoretically estimate drug affinity and/or efficacy based on the molecular geometry/topology of a chemical. Not just something along the lines of "oh the addition of this functional group will increase efficacy" but something more robust and comprehensive that would allow for a somewhat accurate way to calculate affinity and/or efficacy prior to experimental testing. I've been searching, but can't really find too much on PubMed or Google.
-
N&PD Moderators: Skorpio
Geometry and Drug Affinity
- Thread starter saetia95
- Start date
endotropic
Bluelight Crew
- Joined
- Feb 14, 2012
- Messages
- 1,140
This is possible, but requires detailed knowledge of the site you want the chemical to bind at. The best way is to get a crystal structure of the target (receptor, transporter, enzyme, etc.), pick out your binding pocket, then orient chemical groups so that they complement the exposed chemical groups in the binding pocket.
Sadly there's only a few crystal structures for membrane receptors because the fatty membrane makes crystallization a major pain. There still aren't any crystal structures for reuptake transporters like DAT and SERT, and really for most of the targets we would want to dock a drug to.
When crystal structures don't exist the next best option is to look at the amino acid sequence for your target and essentially guess how it would fold into a 3D structure, then dock your drug into that guess.
Failing that you can take a range of chemically related compounds and run in vitro assays to determine their affinity for your target, then use those results to extrapolate how new chemicals would interact. Since we have a lot of different chemicals that bind to the psychoactive drug targets this is probably the easiest and most common kind of SAR that you'll see.
Even with a crystal structure predictions aren't totally accurate. You have to keep in mind that these protein targets are constantly shifting and moving, interacting with other proteins, being chemically modified, etc. There's really no substitute for experimental testing.
sekio
Bluelight Crew
The long and short of it is, no, there's no easy way to go from a molecular structure to an accurate prediction of effects aside from bioassays. Certainly you can try docking ligands to a model of a GPCR. But predicting affinity at a protien target is only part of the problem - you also have the whole Liberation, Absorbtion, Distribution, Metabolism, Excretion, Toxicity panel to work out (ADME) before you can guess at the "potency" of a drug. (c.f. drugs like CB-25 that have high affinity for receptors but poor pharmacokinetics) and there's always the possibility for effects outside of your 'model' - e.g. you can look at a compound like pravadoline in terms of COX affinity, but then you might miss that it is active at CB1 as well, which makes it much more potent than you would "expect".
