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

From: Raik Grünberg <graik.web.de>

Date: Tue, 14 Dec 2004 18:37:18 +0100

Hi there,

I have rather extensively used Amber 8 ptraj for the calculation of entropies

from MD trajectories. With some tricks here and there, things have worked out

quite well - much better (and more efficient) than what I got from my own

feeble attempts of implementing this calculation. Thanks a lot for adding

this method to ptraj!

Unfortunately, I couldn't find any reference for how exactly the entropy

estimate is implemented in ptraj. I haven't got any Fortran knowledge, so I

am pretty lost when looking at the source code.

The classic way of calculating the vibrational entropy from the eigenvalues ev

of the mass-weighted (?) covariance matrix C would be:

(1) S = -1/2 kb SUM( ln ev_i )

or S = -1/2 kb ln | C | [Karplus Kushick 1981]

Then there is the famous Schlitter correction (Schlitter, 1993) which adds

some term to the covariance matrix before determining the eigen values:

(2) C' = C + [ M^-1 h^2 / (kb T e^2) ]

(3) S = -1/2 kb ln | C' |

(M is the diagonal vector of masses)

I suppose the latter is what you are using in ptraj, isn't it?

Next question - in ptraj, the entropy contributions from the decreasing eigen

values neatly converge to zero. In my own implementation any such convergence

was usually messed up by many low but non-zero eigen values and it was never

clear where (or if) I should cut the list. This may of course point to some

mistake on my side - Anyway, is ptraj using any trick to force the

eigenvalues to converge on zero?

Thanks a lot for any help!

Greetings,

Raik

Date: Tue, 14 Dec 2004 18:37:18 +0100

Hi there,

I have rather extensively used Amber 8 ptraj for the calculation of entropies

from MD trajectories. With some tricks here and there, things have worked out

quite well - much better (and more efficient) than what I got from my own

feeble attempts of implementing this calculation. Thanks a lot for adding

this method to ptraj!

Unfortunately, I couldn't find any reference for how exactly the entropy

estimate is implemented in ptraj. I haven't got any Fortran knowledge, so I

am pretty lost when looking at the source code.

The classic way of calculating the vibrational entropy from the eigenvalues ev

of the mass-weighted (?) covariance matrix C would be:

(1) S = -1/2 kb SUM( ln ev_i )

or S = -1/2 kb ln | C | [Karplus Kushick 1981]

Then there is the famous Schlitter correction (Schlitter, 1993) which adds

some term to the covariance matrix before determining the eigen values:

(2) C' = C + [ M^-1 h^2 / (kb T e^2) ]

(3) S = -1/2 kb ln | C' |

(M is the diagonal vector of masses)

I suppose the latter is what you are using in ptraj, isn't it?

Next question - in ptraj, the entropy contributions from the decreasing eigen

values neatly converge to zero. In my own implementation any such convergence

was usually messed up by many low but non-zero eigen values and it was never

clear where (or if) I should cut the list. This may of course point to some

mistake on my side - Anyway, is ptraj using any trick to force the

eigenvalues to converge on zero?

Thanks a lot for any help!

Greetings,

Raik

-- ----------------------------------------------------- Raik Grünberg | Bioinformatique Structurale | Institut Pasteur Tel: +33/1.45.68.87.37 | Paris, France ----------------------------------------------------- ----------------------------------------------------------------------- The AMBER Mail Reflector To post, send mail to amber.scripps.edu To unsubscribe, send "unsubscribe amber" to majordomo.scripps.eduReceived on Tue Dec 14 2004 - 17:53:00 PST

Custom Search