Many thanks to Drs Ross, Carlos, and Yang for the responses about the MD restart.
I do not know why the sander job crashed (Is there a way to check?). I got a kind of error message "pipe broken". (I found it frequently happened after the job runing up to three days). Seems something wrong with the parallel. I run the sander on a linux cluster with 16 dual-nodes (AMD1.6GHz). My MD system includes protein (580 residues) and up to 25000 water. The 2 ns of MD would take almost 2 weeks in general (without competition). Is it normal? (I am using " mpirun -np 32 $AMBERHOME/exe/sander .....). I feel it is kind of slow. Maybe something wrong with the parallel setting, or maybe I need to find an optimzed number of processors (I heard NOT the more the fast)?
Regarding the problem with the rms jump happened with the MD restart, I could not figure out why even I cheched carefully with the input files and settings (irest=1, ntx=7) and looked at the trajectory by VMD. I try to re-run the MD to see if it is happened again.
My target protein is reported as two confomations at the active binding site ("closed" in free and "open" upon ligand binding). Since currently only free closed x-ray structure avaiable, my purpose is trying to catch some "open" conformations in MD simulations and used for subsequent docking and ligand binding study. However, the loop which occupies the active site in free form, bind tightly forming two H-bonds. I did 2ns MD in water also in vaco, and could not find any snapshot that the loop fly out of the active site.Maybe it is a challenge task for Amber. I was thinking to "release" the two H-bonds during the MD thus the loop could be free to move. I noticed Carlos's "locally enchanced sampling", but I am afraid it is really work in this case. My question is, is it possible to achieve such sampling using MD? Does someone have such experiments or know some similar report?
Finally a small question, I got the B-factors (by residues) from MD simulation, and I want to compare it with the experimental B-factors in PDB. Is there a script so I can extract the experimental B-facators easily from PDB and plot it by residues?
Thank you very much for your help!
Xin
---
-------------------------------------
----- Original Message -----
From: Ross Walker
To: amber.scripps.edu
Sent: Monday, October 11, 2004 3:06 PM
Subject: RE: AMBER: Question about restart MD
Dear Xin,
>I was runing a 2ns MD simulations using Amber8. While the job crashed at about 1.5 ns (around NSTEP=750000),
Do you know why it crashed?
>I restarted the job by using the restart coordinate file and giving a new file name for -r (md_new.rst) and
>-x (md_new.mdcrd). The job was restarted at about the 1.5ns (not the exact stop time) but the NSTEP
>started from the begining (NSTEP=100). THe problem is that the job was still running even the time up
>to 2.5 ns (because of the NSTEP). Should I change the "nstlim=1000000" before restarting the simulation
>or some other options I should change to control the restart time?
Yes, sander gets the "time" from the restart file but simply uses the file to obtain the coordinates and velocities. It does not take into account any other settings, time step, number of steps that have occured etc. Hence you need to manually adjust the number of steps in your mdin file to ensure that you stop where you want to. You can work out how many steps are remaining by checking the time in the top of the restart file and dividing by your time step. Then just subtract this from the total number of steps you were hoping to have in the previous run. Note, sander does not write a restrt file on every step (for performance reasons) but instead writes one every ntwr steps. Hence the last integer multiple of ntwr steps from the number of steps actually run will be when the restrt was written.
Although your new job will run for too long, you could always manually terminate it once it has run as far as you want.
>Is there any way to add the restart output (mdcrd,..) to existing files while not creating a new one (mdcrd) ?
You can do this by post processing with ptraj. Just give it each of the mdcrd files, along with the number of frames you want to extract from each one (necessary if ntwx does not equal ntwr). and then as trajout just give it a single mdcrd file. You will then get a single mdcrd file containing the data from each of the input mdcrd files in turn.
>Another question, I mannually terminated the MD job and process the two trajectory files (initial + restart) using ptraj.
>When I plot the rms, I found there is a big change right at the time of 1.5 ns (from 1.2A remarkedly to 2.5A).
>Is it because of the restart (or restart unproperly)?
This depends... Do other things like the energy etc jump? Also, does this jump occur only in the restarted file. I.e. Is it due to a failed restrt or is this jump actually what caused the simulation to crash in the first place? What was the error message when the first simulation crashed?
Also, when you restarted what options did you specify? Make sure you set irest=1 and ntx=5 - if you didn't do this then the velocity information will not have been read from the restart file and so you will have just started out with new randomised velocities.
Try taking a look at the structures in something like vmd and see if you can see anything obviously wrong.
All the best
Ross
/\
\/
|\oss Walker
| Department of Molecular Biology TPC15 |
| The Scripps Research Institute |
| Tel:- +1 858 784 8889 | EMail:- ross.rosswalker.co.uk |
| http://www.rosswalker.co.uk/ | PGP Key available on request |
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Tue Oct 12 2004 - 16:53:00 PDT