AMBER Archive (2006)

Subject: Re: AMBER: error propagation in MM-PBSA binding free energy calculation

From: lhsong (lsong_at_mail.rockefeller.edu)
Date: Fri Apr 28 2006 - 17:37:42 CDT

Hi Juanyin,
Thank you. it is a typo with the formulation of standard error of the mean in my previous email. I wonder how the standard deviation was calculated for some addition terms such as Ggas=Gele+Gint+Gvdw using the error propagation rule? I checked with some results from MM-PBSA, the SD values for Ggas, Gsol, Gtotal seem not calculated from the propagation rule. In addition, the SEtotal for the binding free energies Gtotal, which obtained from SQRT (SE*SEsubtotal + SE*SEentropy), usually up to ~10kcal/mol, is it sound?

Thanks for your help.

Lihong
  ----- Original Message -----
  From: Jianyin Shao
  To: amber_at_scripps.edu
  Sent: Wednesday, April 26, 2006 9:43 PM
  Subject: Re: AMBER: error propagation in MM-PBSA binding free energy calculation

  I am pretty sure that the output STD means standard deviation because I just read the code.
  And I think the standard error of the mean should be STD/SQRT(N), not what you stated.

  Just my 2 cents.

  Jianyin

  On 4/25/06, lhsong <lsong_at_mail.rockefeller.edu> wrote:
    Dear Amber user,
    In the output of MM-PBSA calculations, the STD should be the "standard deviation" or "standard error of the mean energies"? Since I found it is quite confusing in the error report of free energy components in the literation.
    As I understand, the STD is the standard deviation, and the standard error of the mean value (SE) reported in, for example, Dr. Gohlke's paper (J.Comput. Chem, 2004, 25,238) is obtained by SE=SQRT(STD/N), n is the number of snapshots. The errors propagated to the binding free energies should be: SE = SQRT (SE*SEcom + SE*SElig +SE*SErec).
    Apparently, the standard errors for the calculated binding free energies is highly related with the data set N. When using a typical data set of 100-200 snapshots, I found the propagated standard errors of the binding free energies are usually up to 7-10 kcal/mol, comparable to the calculated binding free energy itselt. Is it sound? Also I note that most of published binding free energy calculations did not report the error at all.

    Does someone give me a suggestion on the error calculation for the binding free energies of protein-protein interaction (or a reason not to report)?

    Thank you for your help

    Lihong

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