Software for Structural Comparisons

[Hammilton]

I'm wondering if anyone has or is aware of, software that can be fed a number of files containing the 3D structures of compounds that bind to a certain target and can then come up with a SAR model for compounds which bind to that target?

A while back I had found the structures for as many compounds that inhibit DAT as I could. I then took them, energy minimized, and compared them to each other and measured the distances between important functional groups. I then came up with a SAR model of the requirements for a DAT inhibitor. Then I found out that someone had already done this and published it a while before I had even thought about DAT inhibitors (I think it was published about 5 years prior to that point). I was kind of surprised when I started reading this paper and I see exactly what I come up with; At least I must have done a good job since they were almost identical. This was back when I posted the image comparing cocaine to dimethocaine, and later those to compared to clofencyclan (which I think I actually found when I was looking for stimulant drugs which seemed to have no research to compare to the model).


I'm hoping that software does exist because if it does, I would hope it has some method of considering the affinity of the drug for a given receptor! I mean, even those drugs which have all of the same elements arranged in roughly the same space, they don't overlay 100%, they're all within some acceptable distance. If binding affinities were considered, we could weight the spatial orientation of one structure higher than another. There are all sorts of things you could consider then.

It can be done by hand, I suppose, but I haven't tried yet. This isn't something I've been working on currently, but I saw a thread where someone used one of the overlays I posted and had that thought.

EDIT: if anyone can suggest a way that this could be done manually, please let me know. I have an idea but there's no sense in reinventing the wheel if it's already been done.

Anyway, any thoughts you might have would be appreciated.

EDIT II: Before anyone says "Chembiodraw 3D can do it" - as far as I know CBD does great overlays, but there's no computerized way within it to consider receptor affinity to 3D configuration.

 

[fencamfamine]

I've not actually used it, but I notice that so many researchers refer to 'Discovery 7' in their papers. It seems to deal with the docking and all kinds of things. If you find it's useful, please give a review (for what it's worth, I use Chem Office Suite).

 

[nuke]

Tripos SYBYL-X is a popular one.

http://www.tripos.com/index.php?family=modules,SimplePage,,,&page=SYBYL-X

 

[Hammilton]

That seems to be exactly what I want to use.

Besides opioids and DAT inhibitors, can anyone comment on any other classes of CNS drugs which are particularly amenable to this sort of simulation? I would think GABAergics are diverse to be well suited for this sort of thing. Wrong?

 

[sekio]

Imidazoline receptor
Calicum channel
5-HT2A/C and 5-HT1A serotonergic receptors
Dopamine D2 receptors, esp. autoreceptors
Sigma1 receptor
NMDA receptor
Nicotinic receptors
GHB receptor
Aromatic Hydrocarbon receptor
Cannabinoid receptors (off topic; apparently 2cb-fly has affinity for CB2 in the millimoles)
Nociceptin receptor if you don't consider it an opiate receptor
Fatty acid amide hydrolase
Dopamine / Norepinephrine / Serotonin / Vesicular Monoamine transporters
AMP-activated protien kinase / PPAR delta
Adrenergic receptors, both types
Adenosine receptors

I would think a lot of these would be contingent on accurate secondary/tertiary structures being availiable. Lots of fun drug SAR targets are G-protien coupled receptors and there is a severe lack of structural data for them to the best of my knowledge.

 

[Hammilton]

I'm not asking for a list of CNS receptors.. I was just asking whether there were any other receptors which had a set of ligands that would be amenable to this sort of comparison. DAT isn't overly complicated, but the ability to come up with an optimized configuration based on 3D structure and receptor affinity is interesting. It might make it possible to come up a rough way to predict affinity for more novel structures.

I tried making an image to show what I'm talking about better, but for whatever reason chembiodraw3D is being a dick. With DAT inhibitors you can take a whole bunch of them and you quickly see that they all have a fairly similar 3D configuration- the triangle between the amine, hydrogen bond acceptor and phenyl ring has roughly the same size and shape, to simplify it.

I'm wondering if any other receptors have a family of agonists which lend themselves to this sort of comparison. DAT obviously does, as does sugar (not a CNS receptor of course, but there's a sweetness triangle or something, which describes the required positioning of functional groups involved in binding) but from what I can tell, it's still impossible to do with CB1 which has a very diverse set of agonists.

I'll look into the ones you mentioned, but I don't think there are that many that this will work for.

 

[atara]

That seems to be exactly what I want to use.

Besides opioids and DAT inhibitors, can anyone comment on any other classes of CNS drugs which are particularly amenable to this sort of simulation? I would think GABAergics are diverse to be well suited for this sort of thing. Wrong?

The ligand-gated ion channel type receptors (5-ht3, NMDA, AMPA, kainate, GABA-A, nAChR) all seem to have a lot of binding sites and so you'd have to make sure you're looking at things which bind to the same site.

I think serotonin releasers would be good to study, though. Tramadol, indeloxazine, alpha-ethyltryptamine, MDA, and 5-IAI -- that's a long and weird list. Especially considering that 7-methyl-AET is four times as potent as AET and 5-methyl-MDA is five times as potent as MDA.

 

[Hyperthesis]

"Molecular Operating Environment" (MOE) as well as Accelrys' "Discovery Studio" are both capable of doing what you asked for, Hammilton.
I can't describe the whole procedure here, because it would take me too long, but the basic steps are:
1. Add an arbitrarily long list of 2D-structures to the program (ie. create a database).
2. Minimize these structures with an appropriate method (eg. MMFF94x) so that realistic three-dimensional models result. Often it's useful to keep several poses per structure, which means that your database will eventually grow exponentially. The process of minimization is fully automatic; once you laboriously added all your structures of interest, you chose a method and simply click a button...then wait some hours to days, depending on your computer. Accelrys has for example another program called "Pipeline Pilot", which allows you to pull whole databases of structures through essential steps like energy minimization, calculation of partial charges, calculation of the ionization state, etc...
3. From the now optimized database the aforementioned programs are able to a) dock the structures into a given target, or b) to deviate a pharmacophore model (what you called the "SAR") or even further tasks.

Unfortunately, this isn't thaught/learned in a day or maybe a week. But it is possible for somebody, who is dedicated and willing to spend some time.

As a sidenote, let me point out that it doesn't help to compare structures with differing binding modes resp. binding pockets. It like that apples-oranges-thing... But I guess you knew that already.

Edit: Damn me! Atara beat me to it

 

[Iodjini_dk]

Maestro from Scrödinger software can do calculations like this I guess. But its awfully expensive as far as Im aware. Some computational chemists at the department with which Im affiliated use it for pharmacophore modelling, docking and these sorts of things. The program package allows you to set a lot of constraints including binding affinity and such. But if its for personal use you should probably forget about it unless youre a millionaire