Previous Thread
Next Thread
Print Thread
Basic questions about QUSI
#17421 03/12/08 02:01 PM
Joined: Mar 2007
Posts: 148
J
jeffrey Offline OP
Forum Member
OP Offline
Forum Member
J
Joined: Mar 2007
Posts: 148
Dear all,

I searched the forum for how to use qusi in entropy calculations,but still have some questions unclear. These may be stupid, but I really need to clarify them.

1. Why does one delete hydrogen before qusi? I know one reason is to reduce the storage requirement. Are there any further reasons to do this?

2. What the standard of how long the trajectory will be required for entropy calculation using qusi?

I tested a 5, 4 ns trajectory, including 5000 frames produced by CPT dynamics. The difference is about 0.01 Kcal/mol at 300K. Can it be considered as a convergent results?


Thanks for any suggestions.

------
Jeffry

Last edited by jeffrey; 03/12/08 02:23 PM.
Re: Basic questions about QUSI
jeffrey #17422 03/12/08 03:02 PM
Joined: Nov 2003
Posts: 36
Forum Member
Offline
Forum Member
Joined: Nov 2003
Posts: 36
If you are running an entropy calculation, I would probably leave the hydrogens in place if your computer can handle the storage. The loss of the methyl rotations, hydrogen bending motions, etc should matter theoretically. Whether or not CHARMM will be accurate enough to count for all of it, I am not sure. You can always run it with and without the hydrogens and see how much it matters.

Deletion of the hydrogens is usually an effort to improve convergence, rather than just a storage requirement. You usually need at least 3*Natom snapshots in order to reach convergence. It is an easy way to reduce Natom.

What you are typically concerned with in Quasiharmonic analyses are the large, low frequency motions that do not tend to involve perturbations of the hydrogens. The high energy modes are going to be much less accurate. Furthermore, I think it is fair to say that since you run your MD with the hydrogen bond lengths constrained with SHAKE, you are losing the hydrogen stretching motions anyway.

You'll know it is converged if modes 1 through 6 after QUASi are equal and mode 7 looks like a realistic IR frequency. 1 through 6 are the translational and rotational modes. They should be near 0 after normal mode analysis with the DIAG command, but will be large values after QUASi. Run a testcase as an example.


I wouldn't necessarily feel comfortable with calling a simple 0.1 kcal/mol difference a true convergence there. I'm a bit confused about what you actually compared there. 5000 vs 4000 snapshots? How big is your system? That might not be a meaningful difference in sampling.


Joshua Ward Graduate Student Purdue University Department of Medicinal Chemistry and Pharmacology
Re: Basic questions about QUSI
jmwict13 #17423 03/12/08 05:40 PM
Joined: Mar 2007
Posts: 148
J
jeffrey Offline OP
Forum Member
OP Offline
Forum Member
J
Joined: Mar 2007
Posts: 148
Thanks very much for your detailed explanation.
Now I have a 5ns equilibrated trajectory with a time step of 1fs. The entropy is calculated by using 0-4, 0-5ns, respectively. The two results have a difference of 0.01 kcal/mol at 300 K. The system contains about 4600 atoms, and 1629 non-hydrogen atoms.

I have two more questions:

1. Under what condition the binding free energy can be evaluated only by the trajectory of complex(trajectory of complex, two unbound trajectories seperated from the trajectory of complex)

2. If I use NMA to caculate the entropy, how should I do to overcome the lack of storage requirement by choosing the options of NMA? Is it appropriate to delete the hydrogen atoms, too?


Any advice is greatly appreciated.
Have a nice day.


----
Jeffry

Re: Basic questions about QUSI
jeffrey #17424 03/13/08 10:14 PM
Joined: Sep 2003
Posts: 8,506
rmv Online Content
Forum Member
Online Content
Forum Member
Joined: Sep 2003
Posts: 8,506
Normal mode calcs use energy evaluations, so deleting the H atoms would be a bad idea.

Usually, the first few hundred ps or more of a simulation should be excluded from the analysis.

One way to reduce memory usage is to compute a subset of the normal modes, e.g. the lowest 100, then 101-200, etc.

The DIMB method can also be used to reduce memory.

See vibran.doc ...

Re: Basic questions about QUSI
rmv #17425 03/14/08 12:45 PM
Joined: Mar 2007
Posts: 148
J
jeffrey Offline OP
Forum Member
OP Offline
Forum Member
J
Joined: Mar 2007
Posts: 148
Thank you Rick.
I am trying DIMB.

---
Jeffry

Re: Basic questions about QUSI
jeffrey #17426 05/21/08 04:22 PM
Joined: Jan 2008
Posts: 4
D
Forum Member
Offline
Forum Member
D
Joined: Jan 2008
Posts: 4
Hello everyone,
I have a receptor/ligand complex which I have simulated and calculated the entropy from, using the quasi-harmonic analysis.

The relative entropic change was calculated as:
S(complex) - S(receptor) - S(ligand).

Now I have calculated this using all atoms, non-hydrogen atoms, and only hydrogen atoms.

The magnitude of all values is greater using all atoms, as expected, but the relative changes are also of a greater magnitude with the all atom calculation.

Oddly, the "only-Hydrogen" based calculation was closer to the all-atom then the "non-hydrogen" calculation.

From this, one might be more inclined, when memory is an issue, to use only hydrogens to better estimate relative entropic changes, instead of non-hydrogen atoms only.

If someone could please offer any insight or other contrary anecdotal results, I would be very appreciative.

-danny.

Re: Basic questions about QUSI
dannyH #17427 05/21/08 08:20 PM
Joined: Sep 2003
Posts: 4,797
Likes: 2
Forum Member
Online Content
Forum Member
Joined: Sep 2003
Posts: 4,797
Likes: 2
Qualitatively there are two opposing effects here: Including more atoms usually increases the entropy since there are more conformations available, but when you look at different subsets of similar size there may be an entropy difference due to different amounts of correlation within the subsets. The latter would explain why you find the "only-hydrogen" entropy to be larger - the hydrogen motions are probably less correlated than the motions of the heavy atoms.


Lennart Nilsson
Karolinska Institutet
Stockholm, Sweden

Moderated by  BRBrooks, lennart, rmv 

Link Copied to Clipboard
Powered by UBB.threads™ PHP Forum Software 7.7.4
(Release build 20200307)
Responsive Width:

PHP: 5.6.33-0+deb8u1 Page Time: 0.015s Queries: 29 (0.007s) Memory: 0.9445 MB (Peak: 1.0533 MB) Data Comp: Off Server Time: 2020-10-29 06:57:47 UTC
Valid HTML 5 and Valid CSS