Bond Length from IR Spectroscopic Data

[sol66]

So I'm doing research on a particular protein and I need to find bond lengths to parameterize the chromophore. I currently have a bunch of IR Spectroscopic data and I'm thinking I can use the wavelengths of absorptions between bonds to somehow find bond length. I am actually a physicist and so anything regarding the molecular physics of this would be helpful. Also If I can't calculate bond lengths, I might need to do an "AB Initio QM" calculations of the bonds. Would you have any good sources on how this is done and what software I can use?

 

[Amok]

Take a look at this page:

http://hyperphysics.phy-astr.gsu.edu/hbase/molecule/vibrot.html

Another way of finding bond lengths (I think!) is to have tons of spectroscopy data, find all kind of rotational and vibrational constants (including anharmonicity and rotational vibrational coupling, centrifugal distortion, etc...) using combination differences and use them to describe your potential energy surface (which would be a Taylor series with as many terms as you can afford) and then find it's minimum.

The most popular ab inition calculation programs is probably Gaussian or Gamess, but they are not free. There's another program called ORCA, which is free, but I have no idea on how to use it. But anyway, in anyone of these programs you would just have to find out how to do a geometry optimization.

On another note, it's funny that you mention chromophore and IR spectra at the same time because chromophores absorb in the visible to UV regions (which means electronic transitions).

 

[chemisttree]

For large biomolecules and proteins, http://ambermd.org/" is better... it goes up to eleven!

 

[Amok]

There's also GROMACS now that you mention it. Although I don't think molecular dynamics can be considered ab-inition. Isn'y AMBER a particular force-field anyway?

 

[chemisttree]

I believe it is. It does spit out bond lengths for large molecules rather quickly. He is only parameterizing the chromophore anyway, so MM type calculations would be a good place to start with that.

AMBER has been combined with QM methods for these types of problems.

M.J. Field, P. A. Bash, M. Karplus, "A Combined Quantum-Mechanical and Molecular Mechanical Potential for Molecular Dynamics Simulations", J. Comp. Chem., 1990, 11, 700-733.

 

[sol66]

Ya, well ... I actually have gromacs ... but gromacs requires me to parameterize the bond lengths for my chromophore. My group seems to be using MM for simulating the energy level of the chromophore which seems a bit silly to me since it doesn't take into account quantum mechanics of an sort. Currently I'm using this free program called ABINITI which is very nice. It'll do all the calculations for me for quantum mechanic bond lengths and such. Though I'm thinking to switch from my gromacs forcefield to something else that can do more accurate calculations. I've never done quantum chemistry,... so this is all very new and self taught.

 

[alxm]

I actually have gromacs ... but gromacs requires me to parameterize the bond lengths for my chromophore.

Normally, you can determine these things from IR/microwave spectra of simple molecules. But chromophores can be quite large. I'm doubtful. How would you find and identify the modes of the respective bonds among all those peaks?

My group seems to be using MM for simulating the energy level of the chromophore which seems a bit silly to me since it doesn't take into account quantum mechanics of an sort.

That depends entirely on the context. MM methods are only as good as their parametrization, but when they're used for what they're meant to be used for, they're generally more accurate than most QM methods.
But obviously you can't use an MM method to parametrize itself, and an MM method is not more accurate than the QM method used to calculate the parameters.

But I wouldn't put the study of light-absorption in prosthetic groups to the category of things well-described by MM at all.
What exactly are you (ultimately) trying to calculate? The actual light absorption process?
The difference in conformational energy? (assuming there's a conformational change) etc.

Currently I'm using this free program called ABINITI which is very nice. It'll do all the calculations for me for quantum mechanic bond lengths and such. Though I'm thinking to switch from my gromacs forcefield to something else that can do more accurate calculations. I've never done quantum chemistry,... so this is all very new and self taught.

As a quantum chemist, I can tell.. But I'm obliged to say that QC methods aren't 'black box methods'. Anyone can download a program, read the manual, and use it. But you can't get meaningful results without knowing what you're doing. The software is relatively easy to use, but they don't come with any safety-nets whatsoever.

There are plenty of good textbooks to learn the basics and underlying theory from. But if you haven't worked with it before, I'd suggest you might want to find an experienced quantum chemist to collaborate with before using it in research. There's just a heck of a lot of stuff the books won't teach you. Which method (at the moment!) is the most accurate, but still computationally-feasible, for your problem? What's the error for energy? For geometries? What's a good model size, and what should the model look like? In general: How should you go about doing the calculations, practically? What corrections, interpolations and other additional things might you need to do? Is the result even correct? How do you tell if it converged to the wrong wave function or geometry? Etc. These days, there's a lot of papers being published where, frankly, people simply didn't know what they were doing.

Now, I don't mean to dissuade you! I'm happy to see more people do quantum chemistry. It's just doing quantum chemistry without knowing quantum chemistry that I'm wary of. Bad results are just a waste of time for everyone involved, after all. As I said, QM methods aren't always more accurate than MM methods. (or rather, they're generally less accurate, since MM methods just aren't used, or shouldn't be used, where they're not well-parametrized) As chemisttree notes, you also have QM/MM methods (I've used some myself). If you want to, say, try to reproduce light absorption in the chromophore, followed by some conformational change in chromophore and a larger-scale conformational change in the protein, that would be your only choice (if you could make it work).

If you wanted to calculate the change in energy of the entire protein after a change in the chromophore (skipping the actual absorption process), you could do that with MM alone, but it's very difficult to get an accurate value there. A protein has thousands of degrees of freedom, and correspondingly many local minima on its potential-energy surface. How do you know your energy is being being calculated relative the correct states? (E.g. say an unrelated part of the protein was 'stuck' in a local minimum that's high in energy, and and moved down to a lower one during your optimization, thus giving an apparent drop in energy that's larger than it should be) if you only want to study a conformational change in the chromophore itself, on the other hand, you may well be better off with a pure QM calculation, because there's no problem finding the correct energetic-minimum conformations for a smaller system.

I'm happy to give a suggestion, but it all depends on what you're actually trying to do in the end.

 

[sol66]

Hi alxm, well the problem is that I don't have anyone to really guide me ... other than what is out there on the net. The professor I work under doesn't give me anything to read to help, and if I do ask him for help he tells me to go talk to my co-worker. My co-worker is foreign and extremely difficult to understand. Also, I get the impression that he is new to the subject if he is telling me that there is a possibility that I would have better luck with parameterizing. He also doesn't help much when it comes to advice. Everything that I have done up to this point has been self taught...Anyways, as far as I know, we are doing a normal mode analysis on our chromophore ( which is rather small, only 20 or so carbon skeleton atoms ) which I can't really pretend to tell you what that might imply, other than modeling your chromphore as a bunch of atoms attached to springs and characterizing the oscillations. I think the main objective is to look into the process of how light interacts with retinal, through looking at the difference in energy of the ground state energies and excited state energies. We also look at how this changes the dielectric properties of our atoms in our chromophore. Anyways, as I said ... my main goal is to find correct bond lengths and force constants with my abiniti software so that I can correctly do this modeling. I've also just recently got a book called Electron Structure: Basic Theory and Practical Methods by Richard M. Martin which I'm hoping could aid me with understanding the ABINITI calculations. I've taken a graduate atomic physics class, so I'm hoping that my background should be sufficient. I think for now, I might just parameterize my chromophore using general alkene, alkane, and carbon ring bond distances ... but they will probably skew the results given that the protein probably induces stresses and strains on these bonds that normal organic molecules do not feel.

 

[chemisttree]

You might check your results with those described http://www.springerlink.com/content/dmuxcbtrb78humrv/" when the time comes.

 

[sol66]

Well, I'd probably should just use those results.

 

[chemisttree]

...if I do ask him for help he tells me to go talk to my co-worker. My co-worker is foreign and extremely difficult to understand.

Is your co-worker Italian?

 

[sol66]

At the risk of being caught, I can't say anything about my co-worker. He's a nice guy anyways... just very difficult to understand and doesn't like to explain things.

 

[chemisttree]

I was only joking since the paper was written by Italian scientists.

 

[alxm]

Anyways, as far as I know, we are doing a normal mode analysis on our chromophore ( which is rather small, only 20 or so carbon skeleton atoms ) which I can't really pretend to tell you what that might imply, other than modeling your chromphore as a bunch of atoms attached to springs and characterizing the oscillations. I think the main objective is to look into the process of how light interacts with retinal, through looking at the difference in energy of the ground state energies and excited state energies.

Hmm. But as far as I know, in a chromophore like retinal, the excited state is an electronic excited state. Something like a pi->pi* excitation which allows a bond to rotate and a conformational change. A normal mode analysis would only treat vibrational states, and MM/MD methods can't handle electronically excited states at all.

Neither ab initio methods or MM methods give terribly accurate vibrational states in themselves as far as I know. I have less experience with MM, but with QM methods the results are at best qualitative; you can use it to identify lines in a known spectrum, but not predict them very accurately qualitatively. (There's a large set of extrapolation and error-correction methods for improving those values, as well. At the very least you'd typically scale the frequencies by an empirical constant.)

But it sounds more like you're trying to determine the difference in energy for the chromophore + protein in its different conformations, i.e. after it's absorbed the light, changed conformation and is again in electronic ground-state? As I said before, that's a situation where MM would be required, although it's tricky due to the many local minima.

my main goal is to find correct bond lengths and force constants with my abiniti software so that I can correctly do this modeling.

The software implementation doesn't matter, the method does. Do you know what method you're using and if it's actually any good at what you're trying to calculate? As far as I know, the ABINIT program doesn't implement the hybrid DFT methods that would typically be used for this kind of application, such as B3LYP. Does it do frequency calculations? You need to be able to do that. (And usually, you can't get the force constants directly from the Hessian, it takes a bit of linear algebra; see the paper by Seminario and his FUERZA method for that. I once wrote an implementation of that specifically for getting MM parameters)

I've also just recently got a book called Electron Structure: Basic Theory and Practical Methods by Richard M. Martin which I'm hoping could aid me with understanding the ABINITI calculations.

Looking at the table of contents, I'd recommend you find another book. That one looks oriented towards solid-state calculations, which uses very different methodology and basis sets. (And is conceptually more difficult as well) To suggest some more relevant books that also cover aspects of MM/MD, there's Jensen's "Introduction to computational chemistry", Leach's "Molecular modelling: principles and applications" by Leach, Chris Cramer's book whose title I can't recall offhand, and Levine's "Quantum Chemistry".

 

[sol66]

The program I use recommended that book in the tutorial. You're right, I was a bit disappointed to see that it focused on solid state physics. I'll check out your sources to ... thanks for the advice and help.