- Contemporary messages sorted: [ by date ] [ by thread ] [ by subject ] [ by author ] [ by messages with attachments ]

From: Holger Gohlke <gohlke.bioinformatik.uni-frankfurt.de>

Date: Wed, 05 May 2004 18:32:23 +0200

Marcin Krol schrieb:

*>
*

*> Dear Holger
*

*> Thank you for your answer
*

*>
*

*> I still have a couple of questions:
*

*> 1. I am using MM-PBSA to calculate free energy of binding and I want to
*

*> include entrophy in my calculations. The initial values in the examples
*

*> script perform the minimization of the structure, before nmode is run, to
*

*> DRMS=0.1. I found your e-mail at the amber forum that this is way too much
*

*> and should be changed to DRMS=0.0001 (at least). Then, I have to increase
*

*> MAXCYC as well. The problem is, I am not sure that I get 0.0001 gradient
*

*> at all using Conjugate-Gradient minimization method. I think, it would be
*

*> better to run mode twice after the minimization in sander: first to futher
*

*> minimize the structure in nmode using Newton-Raphson algorithm and then to
*

*> calculate normal modes. Is there an easy way to force mm_pbsa.pl to run
*

*> nmode twice (first for minimizaltion, then for normal modes)? Or maybe
*

*> this is not necessary and conjugate gradient will do its job (I am just
*

*> testing it and increades MAXCYC to 10000 and it looks like it is
*

*> oscillating)
*

Since the NR minimization (in general) involves the calculation and

diagonalization of the Hessian matrix in every step, it is very costly

to apply to larger systems. From my experience, CG minimization led to a

rms value of the gradient < 0.0001 for a 250 residue protein in about

12000 steps. You mention oscillating rms values of the gradient - do you

use a nonbonded cutoff that is larger than your system size?

Best regards

Holger

*>
*

*> Best wishes
*

*> marcin
*

*>
*

*> > > I wonder how the entropy is calculated by ptraj analyze matrix thermo
*

*> > > command. Is this obtained from a quasiharmonic analysis?
*

*> >
*

*> > Yes, the "ptraj analyze matrix thermo" command should give identical
*

*> > results to the quasih program.
*

*> >
*

*> > > What is the
*

*> > > diffrence between entropy calculated in nmode (mm_pbsa.pl script) and
*

*> > > analyze matrix?
*

*> >
*

*> > nmode does a normal mode analysis (i.e. you diagonalize a Hessian matrix
*

*> > obtained from a single conformation residing at a minimum of the
*

*> > potential energy surface), thus you approximate your potential energy
*

*> > surface at that point by a harmonic model.
*

*> >
*

*> > analyze matrix calculates a (mass-weighted) covariance matrix from a
*

*> > trajectory, whereby the inverse of this matrix is related to a "quasi"
*

*> > Hessian matrix. In this "quasiharmonic" analysis, some effects of
*

*> > anharmonic terms in the potentials are included, as are e.g. solvent
*

*> > effects. When calculating entropies by this approach, you need to check
*

*> > the convergence of your results, though.
*

*> >
*

*> > > Could you point me to an article describing how the
*

*> > > entropy in the analyze matrix command is obtained.
*

*> >
*

*> > A good starting point is Curr Opin Struct Biol 1994, 4, 285-290. In
*

*> > particular see ref. 26 and 27 therein.
*

*> >
*

*> > Best regards
*

*> >
*

*> > Holger
*

*> >
*

*> > >
*

*> > > Thank you in advance
*

*> > > marcin
*

*> > >
*

*> > > Dr Marcin Krol
*

*> > > Zaklad Bioinformatyki Collegium Medicum UJ
*

*> > > Kopernika 7E
*

*> > > 31-501 Krakow
*

*> > > tel/fax (012) 422-77-64
*

*> > > e-mail mykrol.cyf-kr.edu.pl
*

*> > > -----------------------------------------------------------------------
*

*> > > The AMBER Mail Reflector
*

*> > > To post, send mail to amber.scripps.edu
*

*> > > To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
*

*> >
*

*> > --
*

*> > ++++++++++++++++++++++++++++++++++++++++++++++++++
*

*> > Dr. Holger Gohlke
*

*> >
*

*> > J.W. Goethe-Universität
*

*> > Fachbereich Biologie und Informatik
*

*> > Institut für Mikrobiologie
*

*> > Marie-Curie-Str. 9
*

*> > 60439 Frankfurt/Main
*

*> > Germany
*

*> >
*

*> > Tel.: (+49) 69-798-29411; Fax: (+49) 69-798-29826
*

*> > Email: gohlke.bioinformatik.uni-frankfurt.de
*

*> > URL: http://www.rz.uni-frankfurt.de/~hgohlke
*

*> > ++++++++++++++++++++++++++++++++++++++++++++++++++
*

*> > -----------------------------------------------------------------------
*

*> > The AMBER Mail Reflector
*

*> > To post, send mail to amber.scripps.edu
*

*> > To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
*

*> >
*

*> -----------------------------------------------------------------------
*

*> The AMBER Mail Reflector
*

*> To post, send mail to amber.scripps.edu
*

*> To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
*

Date: Wed, 05 May 2004 18:32:23 +0200

Marcin Krol schrieb:

Since the NR minimization (in general) involves the calculation and

diagonalization of the Hessian matrix in every step, it is very costly

to apply to larger systems. From my experience, CG minimization led to a

rms value of the gradient < 0.0001 for a 250 residue protein in about

12000 steps. You mention oscillating rms values of the gradient - do you

use a nonbonded cutoff that is larger than your system size?

Best regards

Holger

-- ++++++++++++++++++++++++++++++++++++++++++++++++++ Dr. Holger Gohlke J.W. Goethe-Universität Fachbereich Biologie und Informatik Institut für Mikrobiologie Marie-Curie-Str. 9 60439 Frankfurt/Main Germany Tel.: (+49) 69-798-29411; Fax: (+49) 69-798-29826 Email: gohlke.bioinformatik.uni-frankfurt.de URL: http://www.rz.uni-frankfurt.de/~hgohlke ++++++++++++++++++++++++++++++++++++++++++++++++++ ----------------------------------------------------------------------- The AMBER Mail Reflector To post, send mail to amber.scripps.edu To unsubscribe, send "unsubscribe amber" to majordomo.scripps.eduReceived on Wed May 05 2004 - 17:53:00 PDT

Custom Search