Mr. Kumar,

You can discuss your problem with Mr. D. Marx. He is at RUB as well. He will give you useful information about the subject of your question. He should also advise you on appropriate theoretical approaches balancing between accuracy/compoutational cost.

First, the system is not so large. If it is organics, is about 3 x GABA receptors, for example (54 354 atoms, attachment below). Your theoretical design includes QM/MM, which you have not shown, but treating only 200 atoms of interest. The broadly used QM/MM, however, applied to 200 atoms shall resulte to a great error in the thermochemistry. Furthermore, to this inaccuracy contributes more then one factor.

Of course, you can get these values, using such methods with a combination with other suitable ones. But an error, which to 60 000 atomic system, you found that it does not contribute as much to DG values?! In this context:

(i) What your expect regarding the magnitude of this error? Or how much should be the error of your theoretical prediction of thermochemical data per method used for your system?

Because of if this error cannot affect significantly the properties of interest, then, the computational cost can be reduced even further.

You have written QMMM/FF, but there are more than few variants of theoretical approaches that are combined, in order to reduce the computational effort and to obtain meaningful information about the energetics. If you would like to receive information from explicit traj analysis, then there is very low costs set approaches, operating with your whole system, with different accuracy.

In this context, here also raise other questions:

(ii) If you have probably achieved information about the free Gibbs energy for the entire system of 60 000 atoms with a higher accuracy; then by which experimental method/s you have planned to compare validity of your calculations and theoretical models?.

(iii) Most important question is why you need traj with MD, when the goal is DG? Traj analysis with QM/MM means optimization at each step. You can do MD at low level and then to optimize selected set molecular fragment at higher level using static approaches.

Please note, accounting for the factors contributing to the total energy, that the value of an isolated GABA receptor (18 188 atoms, UFF, 2bg9.pdb initial coordinates) is with magnitude 10E7 kcal.mol-1. The potential energy of 2 x GABA atomic set (or 36 236 atoms) is with magnitude 10E5 kcal.mol-1. For a system of 3 x GABA (or 54 354 atoms) the order of magnitude of the potential energy is also 10E5 kcal.mol-1 (attachment).

The shown data for 1 x GABA atomic set are more accurate. But there is important the order of the magnitude of 2 x GABA and 3 x GABA systems. In other words if your system is a periodic one, for example, even there would not be needed to account for all units (like in the shown example). The energetics is expressed well within the frame of 2 x GABA units (or even 1 x GABA atomic set). Meaning that your system is reduced about 18 000 atoms in the first case.

Thus, I can rephrase the questions above as, what DG values expect you of your system of 60000-atoms per procedure from the available set MM/QM methods; how many would be the error contribution; and how do these expectations fit on the uncertainty of the experimental method/s, which should prove the validity of your theoretical models, in fact?

These comments are for gas-phase analysis. Furthermore, there are without information about the reference experimental method. Because of, depending on it choice there can be posted other comments. For condense phase there are other set comments (eventually).

Towards the size of the output file, then for one trajectory it is 1KB, including the section titles. The information is:

...................................

Trajectory summary

Time (fs) Kinetic (au) Potent (au) Delta E (au) Delta A (h-bar)

0.000000 0.020431 -155.1642219 0.0000000 0.0000000000000000

....................................

However the size of the file and the number trajectories are not linearly connected.

And, as I mentioned above, with Mr. D. Marx in a discussion, you can combine between different methods, these ones providing you information to a balanced accuracy/cost manner under the large set methods.

Ms. Ivanova, 

Thank so much for helpful comments.