Problem with FUNNEL: fps.ld out of boundary

[casalini...@gmail.com]

Dear plumed users and developers,

I am facing a problem with the FUNNEL command in plumed 2.6. 

Simulation crashes, showing the following message: 

Unknown exception:

(exception type: N4PLMD14ExceptionErrorE)

+++ PLUMED error

+++ at Grid.cpp:170, function PLMD::GridBase::index_t

PLMD::GridBase::getIndex(const std::vector<unsigned int>&) const

+++ message follows +++

ERROR: the system is looking for a value outside the grid along the 1 (fps.ld)

index!

My plumed file is the following:

RESTART

WHOLEMOLECULES ENTITY0=1-8617

lig: COM ATOMS=8608,8609,8610,8611,8612,8613,8614,8615,8616,8617

fps: FUNNEL_PS LIGAND=lig REFERENCE=Start.pdb ANCHOR=3231 POINTS=2.79,5.401,5.13,2.707,5.426,5.017

FUNNEL ARG=fps.lp,fps.ld ZCC=1.6 ALPHA=0.60 RCYL=0.1 MINS=-1.0 MAXS=5.0 KAPPA=35000 NBINS=500 NBINZ=500 FILE=BIAS LABEL=funnel WALKERS_MPI

rmsd: RMSD REFERENCE=Start_beta.pdb TYPE=OPTIMAL

d1: DISTANCE ATOMS=8606,8613

d2: DISTANCE ATOMS=5328,8617

METAD ARG=d1,d2 SIGMA=0.01,0.01 HEIGHT=2.0 PACE=500 TEMP=310.0 BIASFACTOR=15.0 GRID_MIN=0.0,0.0 GRID_MAX=6.0,6.0 GRID_SPACING=0.005,0.005 CALC_RCT RCT_USTRIDE=500 LABEL=metad WALKERS_MPI

LOWER_WALLS ARG=fps.lp AT=-0.8 KAPPA=500000 EXP=2 OFFSET=0 LABEL=lwall

UPPER_WALLS ARG=fps.lp AT=4.5 KAPPA=500000 EXP=2 OFFSET=0 LABEL=uwall

UPPER_WALLS ARG=d1 AT=4.0 KAPPA=500000 EXP=2 OFFSET=0 LABEL=uwall_d1

PRINT STRIDE=500 ARG=* FILE=COLVAR

I can understand an error related to fps.lp; I am a little puzzled about this one. Can this be related to how PBC are handled?

I thank you for your support. 

With my best regards,

Tommaso

[孙宝刚]

Hello,  I encountered the same mistake as yours. Have you solved it?  What is the cause of this problem  

[J Yan]

Dear plumed users and developers,

I have exactly he same problem with Funnel-Metadynamics. The simulation failed at the very beginning.  NOPBC and WRAPAROUND are not helping.

I am using Plumed-master/Gromacs-2019.06 and trying to follow the Nature Protocol paper from Dr. Limongelli's group.

From the board, this seems to be common problem with Funnel. Researchers who succeeded might be able to provide great help for those in struggle.

+++ PLUMED error
+++ at Grid.cpp:169, function PLMD::GridBase::index_t

PLMD::GridBase::getIndex(const std::vector<unsigned int>&) const
+++ message follows +++

ERROR: the system is looking for a value outside the grid along the 0 (fps.lp)
index!

Here is my plumed.dat file

# Remember to create empty HILLS and COLVAR files in case you are doing the first run with RESTART option.
# <fill> = necessary an input.
# <?> = necessary a file.
RESTART
ligand: GROUP ATOMS=3233-3250
WHOLEMOLECULES ENTITY0=1-3250
WRAPAROUND ATOMS=ligand AROUND=2480 GROUPBY=18
d1: DISTANCE ATOMS=2480,3233
lig: COM ATOMS=3233,3234,3235,3236,3237,3238,3239,3240,3241,3242,3243,3244,3245,3246,3247,3248,3249,3250
fps: FUNNEL_PS LIGAND=lig REFERENCE=start_apo.pdb ANCHOR=2480 POINTS=4.697,3.385,3.29,2.096,2.441,1.476 NOPBC
rmsd: RMSD REFERENCE=start_apo.pdb TYPE=OPTIMAL
FUNNEL ARG=fps.lp,fps.ld ZCC=2.0 ALPHA=0.45 RCYL=0.1 MINS=1.0 MAXS=3.0 KAPPA=35100 NBINS=500 NBINZ=500 FILE=BIAS LABEL=funnel
# remove BIASFACTOR if you want to perform standard metadynamics
METAD ARG=d1 SIGMA=0.01 HEIGHT=2.0 PACE=500 TEMP=300 BIASFACTOR=20 GRID_MIN=0.0 GRID_MAX=6.0 GRID_WFILE=grid_w.dat GRID_WSTRIDE=250000 LABEL=metad
LOWER_WALLS ARG=fps.lp AT=-0.2 KAPPA=500000 EXP=2 OFFSET=0 LABEL=lwall
UPPER_WALLS ARG=fps.lp AT=2.5 KAPPA=500000 EXP=2 OFFSET=0 LABEL=uwall
UPPER_WALLS ARG=d1 AT=2.7 KAPPA=500000 EXP=2 OFFSET=0 LABEL=distwall
PRINT STRIDE=500 ARG=* FILE=COLVAR

Best regards and thanks,

Yan

[Robert Coffman]

Something that ended up helping me was increasing the system size. For example, I originally solvated with 15 A solvent around my protein with the following funnel parameters and wall restraints:

FUNNEL ARG=fps.lp,fps.ld ZCC=5.5 ALPHA=0.3 RCYL=0.2 MINS=-0.4 MAXS=6.0
LOWER_WALLS ARG=fps.lp AT=-0.2 KAPPA=3500 EXP=2 OFFSET=0 LABEL=lwall
UPPER_WALLS ARG=fps.lp AT=5.9 KAPPA=3500 EXP=2 OFFSET=0 LABEL=uwall

I then took a shot in the dark and resolvated my system with 35 A solvent around my protein and fixed it crashing at the beginning. I came to the conclusion that with a ZCC of 55 angstroms and a box that couldn't hold it caused my problems. 

Then I would get a crash partway through the simulation. What finally fixed the new crash was changing the restraining wall to be 2 angstrom from the funnel ends instead of 1 angstrom. Essentially changing :

UPPER_WALLS ARG=fps.lp AT=5.9 KAPPA=3500 EXP=2 OFFSET=0 LABEL=uwall

to

UPPER_WALLS ARG=fps.lp AT=5.8 KAPPA=3500 EXP=2 OFFSET=0 LABEL=uwall

I hope this helps some people in the future.

Regards,

Robert

[Tommaso Casalini]

Hi Robert, 

thanks for your message.

I would say that the availability of the FUNNEL module is great, but I fear that it is not fully optimized.

Moreover, it seems that the main developer of the module does not usually check this group, and this makes solving such issues more difficult. 

In my experience, there are still some cases in which there is an unlucky combination of system size/funnel size and position/PBC/protein moving partially out of the box that leads to error messages at least for long simulations. For example, in my case I restrained three atoms of the protein very far from the binding site (I assumed bona fide that this does not play a role) and the simulations run smoothly, while they crashed otherwise. 

In my opinion, the best way to carry out FM is to employ the approach proposed by the authors in the Nature Protocol paper, i.e., running (relatively) short simulations with many replicas using the multiple walker feature (probably it is better if every replica starts from a different system configuration in terms of CV) so that the protein does not appreciably diffuse out of the box and PBC are not an issue. 

Otherwise it may be possible to put back the protein in the center of the box every X ns and modify other inputs accordingly (more tedious but probably the entire operation can be done by a script). 

Best,

Tommaso

--
You received this message because you are subscribed to the Google Groups "PLUMED users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to plumed-users...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/plumed-users/c884b1d4-6872-426f-8af5-982330725acbn%40googlegroups.com.