AMBER Archive (2005)

Subject: Re: AMBER: MM_PBSA_parameter

From: Ray Luo (rluo_at_uci.edu)
Date: Wed Nov 23 2005 - 15:48:17 CST

>Then I ran MM_PBSA for my system. Due to the
>large size of my system, I have to decrease the "SCALE" to 1.5. The topology
>file I use is generate by tleap with default parameters. Total 150 snapshots
>extracted from 3ns trajectory. The GB gives a negative value with small SD. But
>PB gives a large positive value with large SD. I do not know what is wrong.
>Then I tried to change some parameters. I set "PRBRAD" to 1.5 or 1.4. I got
>very different answers. So my question is how to choose "PRBRAD". What does it
>exactly mean to "use 1.4 with the PARSE parameter set and 1.6 with the radii
>optimized by R. Luo"? What is PARSE parameter set and what is R. Luo parameter
>set?
>
>

On the larger SD for PB, have you patched the last pbsa bugfix? If you
have, this is most likely due to the larger grid spacing that you used,
i.e. a smaller SCALE value. Also, could you show us your pbsa input and
output files? There is no enough information for us to tell what is
happening.

If you use RADIOPT of 1 for pbsa, then PRBRAD of 1.6 should be used
because this is the value in the PB cavity radii optimization performed
by our group. If you use the parse cavity radii set, you need to use
1.4, again because this is the valued used in the parse cavity radii
optimization performed by Kim Sharp.

It's not very surprising that PB/GB results are highly dependent on what
radii are used. There is nothing wrong about it just as molecular
dynamics simulations strongly depend on van der Waals parameters,
charges and other parameters. Unfortunately, there are no overwhelming
evidences as for which set of radii are better with the Amber charges
because there are not enough data to constrain the value of PB/GB cavity
radii. This is still an ongoing research in many groups.

Further, PB results also depend on PRBRAD, i.e. water probe radius,
which is used to determine where water molecules can go or not in a
complex molecular environment. GB's dependent on PRBRAD is not visible
to end users because it's used in its parameter optimization process.
However, due to cancellation of errors, relative binding free energies
in general depend less on these parameters than absolute binding free
energies, and much less than conformational free energies.

>Another question is how I can make sure the solution of PB already converges if
>I only set "LINIT" to certain number. Why is the "SURFTEN" different for PB and
>GB? Why do we calculte the absolute reaction field energy ("REFE"=0) instead of
>the difference of reaction field energy ("REFE">0) to calculate the binding
>free energy?
>
>
>
You can find out whether PB has converged by looking at the pbsa output
file. This might be another reason that you saw larger SD for PB
results. There should be a warning on convergence failure in the output
file if PB doesn't converge. Also, if you set "npbverb=1" in your pbsa
input, you can find out more info on your pbsa run.

Finally, the SURFTEN value for PB is used when the PB cavity radii set
(RADIOPT=1) was optimized by our group. I think the SURFTEN for GB is
taken from the parse radii set. However, the difference in SURFTEN won't
change your result noticeably because the difference is very small.

All the best,
Ray 

-- 
====================================================
Ray Luo, Ph.D.
Department of Molecular Biology and Biochemistry
University of California, Irvine, CA 92697-3900
Office: (949)824-9528         Lab: (949)824-9562
Fax: (949)824-8551          e-mail: rluo_at_uci.edu
Home page: http://rayl0.bio.uci.edu/rayl/
====================================================

-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu