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

From: Naser Alijabbari <na3m.virginia.edu>

Date: Tue, 12 Aug 2008 00:17:40 -0400

Hello,

Someone else in my group has written the code to take the output of ptraj

eigen frequencies and calculate absorption coefficients. Feel free to answer

what you have time for and thanks before hand.

My questions:

1) http://amber.ch.ic.ac.uk/archive/200412/0026.html is a discussion about

the difference btw Langevin vs Berendsen temperature control. The end

conclusion is that Langevin corrupts the fast dynamics of the system so it

shouldn't be used during production run, and if one wants to compute time

correlation function then they need to avoid ntt=3. Can anyone elaborate on

this? From the tutorial I was under the impression that Berendsen algorithm

had to be much more elaborate, so is my last step correct? My main focus is

picking up vibrations occurring in a time frame of ~6ps, is it a good idea

to use ntt=3 at all for this type simulation?

100ps simulation of the protein 2TRX w/ 108 residues, AMBER 8

tleap:

set default Dielectric constant

TIP3PBOX w/ 8 angstrom water

I get the following warning because I don't add Na+ ion "WARNING: The

unperturbed charge of the unit: -4.000000 is not zero". I neutralize DNA but

I am not sure if the guideline applied to everything (does it cause problems

with periodic boundary condition?)

Step1:

imin=1

maxcycle=1000, ncyc = 500,

ntx=1, (default)

ntpr=100 (default is 50 for writing to mdinfo)

ntr= 1,

ntb=1

cut =10.0,

restraint_wt=5.0,

restraintmask=':1-108'

Step2:

imin = 1,

maxcyc = 2500, ncyc = 1000,

ntb = 1,

ntr = 0,

cut = 10

Step 3:

imin=0,

irest=0, ntx=1,

ntwr=1000, ntwx=100, ntpr=100,

ntr= 1,

ntslim=8000, dt=.002,

ntt=3, gamma_ln = 1.0, temp0=300, tempi=0,

ntb=1,

ntc=2, ntf=2, cut =10.0,

restraint_wt=10.0,

restraintmask=':1-108'

Step 4:

imin=0,

irest=1, ntx=5

ntwr=1000, ntwx=1000, ntpr=100,

ntr=0

ntslim=20000, dt=.002,

ntt=3, gamma_ln=1, temp0=300, tempi=300,

ntb=2, pres0=1, ntp=1, taup=2

ntc=2, ntf=2, cut=10

Step 5:

imin=0,

irest=1, ntx=5,

ntwr=1000, ntwx=1000, ntpr=100,

ntr=0

ntslim=30000, dt=.002,

ntt=3, gamma_ln=1, temp0=300, tempi=300,

ntb=2, pres0=1, ntp=1, taup=2,

ntc=2, ntf=2, cut=10,

Production run:

imin=0,

irest=1, ntx=5,

ntwr=1000, ntwx=25, ntpr=100,

ntr=0

ntslim=50000, dt=.002, ncsm=1

ntt=1, temp0=300, tempi=300,

ntb=2, pres0=1, ntp=1, taup=2,

ntc=2, ntf=2, cut=10

ptraj:

trajin bbprotein.mdcrd.gz

strip :WAT

trajout bbproteinNOwat.mdcrd

rms first out bbprotein.rms .C,CA,N time .05

matrix mwcovar name bbprotein.matrix out bbprotein.matrix

analyze matrix bbprotein.matrix out bbprotein.mode vecs 400

I am moving the center of mass at every step based on someone else's

example, and I am not sure if it makes a difference to do this during

simulation or in my ptraj. There is a discussion wrt to "pca analysis"

http://amber.ch.ic.ac.uk/archive/200601/0124.html and I am not sure if it

applies. Center of mass was also discussed here

http://amber.ch.ic.ac.uk/archive/200704/0202.html and it suggests that it is

not necessary to do this if one is using ntt=1. Any hints?

2) http://archive.ambermd.org/200603/0110.html and

http://www.rosswalker.co.uk/tutorials/amber_workshop/Tutorial_eight/ says

that if one averages the results and the simulation is long enough two

machines or different number of processors will give the same result. What

is averaged? And what is long enough? Is there a rule of thumb?

3) http://archive.ambermd.org/200805/0518.html is a discussion about

doing multistep production runs (i.e. doing a simulation from 0 to 100ps and

feeding the .rst file to the input of another simulation from 100 to 200ps

or a 200ps in two 100ps steps). Discussion points out that during

calculation higher precision is used than what is written to the .rst file.

This truncation doesn't seem to shift the position of the absorption peaks

for me. However, there is a significant difference between doing a 0 to

200ps run vs 0 to 100ps fed to 100 to 200ps. So a multi step simulation is

less accurate?

4) http://archive.ambermd.org/200407/0148.html is a

discussion about regularity of writing to mdcrd during production run. As I

mentioned before I am interested in the molecular vibrations within ~6ps

time frame. Does this mean that at max. I can get away w/ writing to mdcrd

every 3000 step (ntwx=3000, dt=.002)? Also does it make a different what

value one uses for vecs? I picked a value from an example but I am not sure

what the preference is.

I can't seem to make the absorption coefficients vs cm^-1 plot to settle on

any value since position of absorption peaks change w/ number of processors,

time duration, steps in simulation.

Thanks

Naser

-----------------------------------------------------------------------

The AMBER Mail Reflector

To post, send mail to amber.scripps.edu

To unsubscribe, send "unsubscribe amber" (in the *body* of the email)

to majordomo.scripps.edu

Received on Wed Aug 13 2008 - 06:07:33 PDT

Date: Tue, 12 Aug 2008 00:17:40 -0400

Hello,

Someone else in my group has written the code to take the output of ptraj

eigen frequencies and calculate absorption coefficients. Feel free to answer

what you have time for and thanks before hand.

My questions:

1) http://amber.ch.ic.ac.uk/archive/200412/0026.html is a discussion about

the difference btw Langevin vs Berendsen temperature control. The end

conclusion is that Langevin corrupts the fast dynamics of the system so it

shouldn't be used during production run, and if one wants to compute time

correlation function then they need to avoid ntt=3. Can anyone elaborate on

this? From the tutorial I was under the impression that Berendsen algorithm

had to be much more elaborate, so is my last step correct? My main focus is

picking up vibrations occurring in a time frame of ~6ps, is it a good idea

to use ntt=3 at all for this type simulation?

100ps simulation of the protein 2TRX w/ 108 residues, AMBER 8

tleap:

set default Dielectric constant

TIP3PBOX w/ 8 angstrom water

I get the following warning because I don't add Na+ ion "WARNING: The

unperturbed charge of the unit: -4.000000 is not zero". I neutralize DNA but

I am not sure if the guideline applied to everything (does it cause problems

with periodic boundary condition?)

Step1:

imin=1

maxcycle=1000, ncyc = 500,

ntx=1, (default)

ntpr=100 (default is 50 for writing to mdinfo)

ntr= 1,

ntb=1

cut =10.0,

restraint_wt=5.0,

restraintmask=':1-108'

Step2:

imin = 1,

maxcyc = 2500, ncyc = 1000,

ntb = 1,

ntr = 0,

cut = 10

Step 3:

imin=0,

irest=0, ntx=1,

ntwr=1000, ntwx=100, ntpr=100,

ntr= 1,

ntslim=8000, dt=.002,

ntt=3, gamma_ln = 1.0, temp0=300, tempi=0,

ntb=1,

ntc=2, ntf=2, cut =10.0,

restraint_wt=10.0,

restraintmask=':1-108'

Step 4:

imin=0,

irest=1, ntx=5

ntwr=1000, ntwx=1000, ntpr=100,

ntr=0

ntslim=20000, dt=.002,

ntt=3, gamma_ln=1, temp0=300, tempi=300,

ntb=2, pres0=1, ntp=1, taup=2

ntc=2, ntf=2, cut=10

Step 5:

imin=0,

irest=1, ntx=5,

ntwr=1000, ntwx=1000, ntpr=100,

ntr=0

ntslim=30000, dt=.002,

ntt=3, gamma_ln=1, temp0=300, tempi=300,

ntb=2, pres0=1, ntp=1, taup=2,

ntc=2, ntf=2, cut=10,

Production run:

imin=0,

irest=1, ntx=5,

ntwr=1000, ntwx=25, ntpr=100,

ntr=0

ntslim=50000, dt=.002, ncsm=1

ntt=1, temp0=300, tempi=300,

ntb=2, pres0=1, ntp=1, taup=2,

ntc=2, ntf=2, cut=10

ptraj:

trajin bbprotein.mdcrd.gz

strip :WAT

trajout bbproteinNOwat.mdcrd

rms first out bbprotein.rms .C,CA,N time .05

matrix mwcovar name bbprotein.matrix out bbprotein.matrix

analyze matrix bbprotein.matrix out bbprotein.mode vecs 400

I am moving the center of mass at every step based on someone else's

example, and I am not sure if it makes a difference to do this during

simulation or in my ptraj. There is a discussion wrt to "pca analysis"

http://amber.ch.ic.ac.uk/archive/200601/0124.html and I am not sure if it

applies. Center of mass was also discussed here

http://amber.ch.ic.ac.uk/archive/200704/0202.html and it suggests that it is

not necessary to do this if one is using ntt=1. Any hints?

2) http://archive.ambermd.org/200603/0110.html and

http://www.rosswalker.co.uk/tutorials/amber_workshop/Tutorial_eight/ says

that if one averages the results and the simulation is long enough two

machines or different number of processors will give the same result. What

is averaged? And what is long enough? Is there a rule of thumb?

3) http://archive.ambermd.org/200805/0518.html is a discussion about

doing multistep production runs (i.e. doing a simulation from 0 to 100ps and

feeding the .rst file to the input of another simulation from 100 to 200ps

or a 200ps in two 100ps steps). Discussion points out that during

calculation higher precision is used than what is written to the .rst file.

This truncation doesn't seem to shift the position of the absorption peaks

for me. However, there is a significant difference between doing a 0 to

200ps run vs 0 to 100ps fed to 100 to 200ps. So a multi step simulation is

less accurate?

4) http://archive.ambermd.org/200407/0148.html is a

discussion about regularity of writing to mdcrd during production run. As I

mentioned before I am interested in the molecular vibrations within ~6ps

time frame. Does this mean that at max. I can get away w/ writing to mdcrd

every 3000 step (ntwx=3000, dt=.002)? Also does it make a different what

value one uses for vecs? I picked a value from an example but I am not sure

what the preference is.

I can't seem to make the absorption coefficients vs cm^-1 plot to settle on

any value since position of absorption peaks change w/ number of processors,

time duration, steps in simulation.

Thanks

Naser

-----------------------------------------------------------------------

The AMBER Mail Reflector

To post, send mail to amber.scripps.edu

To unsubscribe, send "unsubscribe amber" (in the *body* of the email)

to majordomo.scripps.edu

Received on Wed Aug 13 2008 - 06:07:33 PDT

Custom Search