Re: [GaMD-discuss] error with amber24 GaMD-PPI Too many TI atoms--current limit: 500 atoms

SourceForge Headquarters 1320 Columbia Street Suite 310 San Diego, CA 92101 +1 (858) 422-6466

Hi Christian,

Great that your system runs with PPI-GaMD by “only modifying the MaxNumberTIAtom”. Thanks for confirming that!

Note that timask1 and scmask1 use atom masks. Anything in the ambmask format should work; it doesn’t have to be defined by residues.
You can find detailed descriptions of PPI-GaMD simulation parameters in the paper SI: https://pubs.acs.org/doi/10.1021/acs.jctc.1c00974.

Victor can probably comment more on your questions as well.

Best,
Yinglong

On Nov 24, 2025, at 7:06 PM, Christian Seitz <cgs...@gm...> wrote:

Hi Victor and Yinglong,

Victor, thanks for this information. I have a few more questions. For clarity, let's say that protein 1 is a small antibody, protein 2 is a large glycoprotein, and I want to see association/dissociation between the two.

1) You are saying that I should select the interfacial residues of protein 1, and then all the atoms of protein 2? I was under the impression that I should select all residues of protein 1, because the GaMD-AMBER manual defines "timask1" as "Specifies atoms of the first (bound) ligand or peptide in ambmask format...". A similar definition is given for "scmask1". But this should be just the interfacial residues, not the entire antibody? And they should be given as residues, not atoms? If so, perhaps the manual could be revised for clarity.
2) I ran x ns of a GaMD equilibration, and am ready to run another part. Thus I should subtract x ns from the overall GaMD equilibration run, as that is already complete. Do I subtract x ns from the normal MD equilibration (ntcmdprep)? Or the biasing MD equilibration (ntebprep), or perhaps something else? Overall, does the equilibration go from ntcmdprep -> ntebprep -> ntcmd -> nteb, or what order of steps does the equilibration run through? This will let me know how to modify my equilibration input files for restarted equilibration runs.

Yinglong, good catch from Jinan. I recompiled amber24, only modifying the MaxNumberTIAtom = 750 -> 5250, and my system runs with this. Contrary to answers I've seen elsewhere, I can leave MaxNumberTIPair and MaxNumberREAFAtom at their default values and my system will run GaMD-PPI. In my GaMD equilibration run, at least, my system does not run any faster by leaving MaxNumberTIPair and MaxNumberREAFAtom at their default values.

Thanks for your help!

Best,
Christian

On Sun, Nov 23, 2025 at 9:11 AM Miao, Yinglong <Yin...@me...<mailto:Yin...@me...>> wrote:
Hi Christian,

I’m glad that your new simulation works. Jinan mentioned that the “essential" code change in Amber for running PPI-GaMD would be increase in “MaxNumberTIAtom", in your case:

MaxNumberTIAtom = 750 -> 5250

Could you do us a favor to double check if that’s “sufficient", and other changes are not necessary?

Thanks,
Yinglong

On Nov 18, 2025, at 9:25 AM, Adediwura, Victor Ayo <vad...@em...<mailto:vad...@em...>> wrote:

Hi Christian,

Glad to hear that the code works!

A few comments, the timask1 and scmask1 should be residue number rather than atom numbers. timask1 and scmask1 are the region you would like to see boosted (e.g for protein-protein dissociation, this could be the interfacial residues of the first protein).  The bgpro2atm and edpro2atm should be the first atom number of the second protein, and edpro2atm is the last atom of the second protein (which is also boosted). Thus, for the first protein, you will have to select interface residues for boosting while for the second protein, you would have to select from the first to the last atom for boosting. In essence, you might want to renumber your PDB such that the first protein where you are selecting some part of the protein (i.e interfacial residues) should come last and the second protein would come first

For the second part, I would suggest to run longer equilibration considering size of your system which is quite large. For how it will play out in this case, I think this is a good idea " If I were to break up one long GaMD equilibration, how would I do so? Just change restart parameters, i.e. irest=1, ntx=5, as if I were restarting a normal, production MD?"

I hope this helps!

Kind regards,

Victor

________________________________
From: Christian Seitz <cgs...@gm...<mailto:cgs...@gm...>>
Sent: Monday, November 17, 2025 6:18 PM
To: Adediwura, Victor Ayo <vad...@em...<mailto:vad...@em...>>
Cc: Miao, Yinglong <Yin...@me...<mailto:Yin...@me...>>; gamd-discuss <gam...@li...<mailto:gam...@li...>>; Davis, James J. <jj...@an...<mailto:jj...@an...>>; Ramanathan, Arvind <ram...@an...<mailto:ram...@an...>>
Subject: Re: [GaMD-discuss] error with amber24 GaMD-PPI Too many TI atoms--current limit: 500 atoms

Hi Victor,

Fantastic, this works! The key change was involving all atoms in the TI mask, not excluding hydrogens:

timask1 = ':1870-2126!@H' -> timask1 = ':1870-2126'
scmask1 = ':1870-2126!@H' -> scmask1 = ':1870-2126'

I would have thought that including the hydrogens in the TI mask would increase the atom limit of the system, and thus push it closer to amber24's TI atom limit. For my own understanding, do you know why including hydrogens in the TI mask allowed the system to run?

Separately, I do not think I can finish a full GaMD equilibration for each of my systems within my computing resource's wallclock limit. Would it be better to run multiple short GaMD equilibrations until the k0P and k0D stabilize at 1.0, or should I run one longer GaMD equilibration and break it up into multiple runs? If I were to break up one long GaMD equilibration, how would I do so? Just change restart parameters, i.e. irest=1, ntx=5, as if I were restarting a normal, production MD?

Many thanks for your help in this.

Best,
Christian

On Sat, Nov 15, 2025 at 12:00 PM Adediwura, Victor Ayo <vad...@em...<mailto:vad...@em...>> wrote:
Hi Christian,

I just checked the files and noticed there is an issue with the equil_gamd.in<http://equil_gamd.in/> file. I have modified this file, and it should work. Pls find attached the modify file.

Kind regards,

Victor

________________________________
From: Miao, Yinglong via GaMD-discuss <gam...@li...<mailto:gam...@li...>>
Sent: Monday, November 10, 2025 8:48 PM
To: Christian Seitz <cgs...@gm...<mailto:cgs...@gm...>>
Cc: gamd-discuss <gam...@li...<mailto:gam...@li...>>; Davis, James J. <jj...@an...<mailto:jj...@an...>>; Ramanathan, Arvind <ram...@an...<mailto:ram...@an...>>
Subject: Re: [GaMD-discuss] error with amber24 GaMD-PPI Too many TI atoms--current limit: 500 atoms

Hi Christian,

It might be better to select all atoms in the protein 1 for timask1 and scmask1.

Other than that, what you changed especially under 2) should have resolved the issue. Could you possibly share your simulation input files and the mdout file? We’d like to run it with our local Amber and see how it works.

Best,
Yinglong

On Nov 10, 2025, at 4:07 PM, Christian Seitz <cgs...@gm...<mailto:cgs...@gm...>> wrote:

Hi Yinglong,

Thanks for the suggestions. I have tried several things so far, let me recap them. My system contains 1,951 heavy atoms (3,814 atoms total) in the TI region, which I am calling through timask1 = ':1870-2126!@H' and scmask1 = ':1870-2126!@H' in an input file to run GaMD-PPI.

1) I compiled a standard, unmodified build of amber24. When I put a large system into that build to run GaMD-PPI, I get an error saying that the TI atom limit is 500 atoms. Thus I need to expand the TI atom limit.

2) I tried changing the "MaxNumberTIAtom", "MaxNumberTIPair", and "MaxNumberREAFAtom" within $AMBERHOME/amber/pmemd24_src/src/pmemd/src/gti_controlVariable.i. I increased these atom limits as shown below, recompiled a fresh version of amber24, and still found the same error, saying the atom limit in TI was 500.

MaxNumberTIAtom = 750 -> 5250
MaxNumberTIPair = 300 -> 3000
MaxNumberREAFAtom = 500 -> 4000

3) The "MaxNumberTIAtom2" within $AMBERHOME/amber/pmemd24_src/src/pmemd/src//cuda/gti_NBList_kernels.cu comes with a limit of 4000, which should be enough for this system.

static const int MaxNumberTIAtom2 = 4000

I did not try increasing that.

Do you have any idea what might be overwriting these other variables to enforce a smaller TI atom limit? I'm thinking the error message, stating there's a TI atom limit of 500, is probably not tied to the actual TI atom limit - I'm guessing that if one made the TI atom limit to be 2 atoms, the error message would still say the TI atom limit is 500. But I cannot figure out what variable is actually controlling the TI atom limit, or how to increase that. Or perhaps there's some issue with how I'm telling GaMD-PPI to select my TI atoms, and it's selecting more than I want? I can send my input files or system files if that helps, or can try any other ideas you have to get this to run. Thanks for your help!

Best,
Christian

On Mon, Nov 10, 2025 at 6:11 AM Miao, Yinglong <Yin...@me...<mailto:Yin...@me...>> wrote:
Hi Christian,

The related parameters should be defined in file $AMBERHOME/src/pmemd/src/gti_controlVariable.i. I compared this file in our "gamd" branch with master branch in the Amber git repository:

# AMBER “gamd"

  STATIC_CONST_INT MaxNumberTIAtom = 4000;
  STATIC_CONST_INT MaxNumberTIPair = 3000;
  STATIC_CONST_INT MaxNumberNBPerAtom = 3072;

  STATIC_CONST_INT MaxNumberREAFAtom = 5000;
  STATIC_CONST_INT MaxNumberRMSDAtom = 200;
  STATIC_CONST_INT MaxNumberRMSDRegion = 5;

# AMBER master

  STATIC_CONST_INT MaxNumberTIAtom = 750;
  STATIC_CONST_INT MaxNumberTIPair = 300;
  STATIC_CONST_INT MaxNumberNBPerAtom = 3072;

  STATIC_CONST_INT MaxNumberREAFAtom = 500;
  STATIC_CONST_INT MaxNumberRMSDAtom = 200;
  STATIC_CONST_INT MaxNumberRMSDRegion = 5;

Three parameters “MaxNumberTIAtom”, “MaxNumberTIPair” and “MaxNumberREAFAtom” are modified. Could you help to try those values as used in the “gamd” branch to see if that resolves the issue?

Thanks,
Yinglong

On Nov 9, 2025, at 11:44 AM, Miao, Yinglong <Yin...@me...<mailto:Yin...@me...>> wrote:

Hi Christian,

Seems you are changing these variables in the wrong direction. You need to increase MaxNumberTIAtom and MaxNumberTIAtom2 to a number greater than 500, like 4000 or the number of atoms in your protein 1 for simulation.

Hope this helps,
Yinglong

On Nov 5, 2025, at 5:16 PM, Christian Seitz <cgs...@gm...<mailto:cgs...@gm...>> wrote:

Hi Yinglong,

Thanks, as a test I reduced the limit in $AMBERHOME/amber/pmemd24_src/src/pmemd/src//cuda/gti_NBList_kernels.cu from 4000 to 400:

static const int MaxNumberTIAtom2 = 4000 -> 400

After building a fresh install of amber24 with this reduced atom limit, I found that the TI atom limit remained at 500. If it helps, inside $AMBERHOME/amber/pmemd24_src/src/pmemd/src/cuda/gti_gpuContext.cpp one can find this snippet:

 if (TISize > gti_simulationConst::MaxNumberTIAtom) {
          std::string e("Too many TI atoms--current limit: 500 atoms");
          throw std::runtime_error(e);

I can only find TISize in $AMBERHOME/amber/pmemd24_src/src/pmemd/src/cuda/gti_gpuContext.cpp. I have tried grepping a variety of terms within $AMBERHOME/amber/pmemd24_src/src/pmemd/src, such as "500", "TIAtoms", and "TISize", but nothing amongst these stand out to my non-expert eyes to change. When I ask chatgpt or claude how to solve this, they simply refer to the forum posts in my initial email, including this post, and suggest searching through non-existent files. Do you have any other ideas on where to search or what to look for? Thanks for your help!

Best,
Christian

On Mon, Nov 3, 2025 at 7:57 PM Miao, Yinglong <Yin...@me...<mailto:Yin...@me...>> wrote:
Hi Christian,

This is what I found for “MaxNumberTIAtom” in the pmemd src code:

~/amber-git/src/pmemd/src$ grep -iR "MaxNumberTIAtom" *
cuda/gti_NBList_kernels.cu:  __shared__ volatile float coordTI[gti_simulationConst::MaxNumberTIAtom][3];
cuda/gti_NBList_kernels.cu:  __shared__ float coordTI[gti_simulationConst::MaxNumberTIAtom][3];
cuda/gti_NBList_kernels.cu:  __shared__ volatile float coordTI[gti_simulationConst::MaxNumberTIAtom][3];
cuda/gti_NBList_kernels.cu:  static const int MaxNumberTIAtom2 = 4000;
cuda/gti_NBList_kernels.cu://  __shared__ volatile bool bondMask[(gti_simulationConst::MaxNumberTIAtom2 + 1) *
cuda/gti_NBList_kernels.cu://                                    gti_simulationConst::MaxNumberTIAtom2];
cuda/gti_NBList_kernels.cu://  __shared__ volatile float coordTI[gti_simulationConst::MaxNumberTIAtom2][3];
cuda/gti_NBList_kernels.cu:    __shared__ volatile bool bondMask[(MaxNumberTIAtom2 + 1) *
cuda/gti_NBList_kernels.cu:                                              MaxNumberTIAtom2];
cuda/gti_NBList_kernels.cu:    __shared__ volatile float coordTI[MaxNumberTIAtom2][3];
cuda/gti_gpuContext.cpp:  uint4 *temp = new uint4[gti_simulationConst::MaxNumberTIAtom];
cuda/gti_gpuContext.cpp:  unsigned* tempSC = new unsigned[gti_simulationConst::MaxNumberTIAtom*2];
cuda/gti_gpuContext.cpp:  unsigned tempShift = gti_simulationConst::MaxNumberTIAtom;
cuda/gti_gpuContext.cpp:        if (TISize > gti_simulationConst::MaxNumberTIAtom) {
cuda/gti_gpuContext.cpp:  int tempList[gti_simulationConst::MaxNumberTIAtom];
gti_controlVariable.i:  STATIC_CONST_INT MaxNumberTIAtom = 4000;
ti.F90:  !!integer, parameter::MaxNumberTIAtom=500 !! this must be consistent w/ gti_controlVariable.i

Maybe you can check MaxNumberTIAtom2 in cuda/gti_NBList_kernels.cu as well?

Best,
Yinglong

On Oct 31, 2025, at 6:28 PM, Christian Seitz <cgs...@gm...<mailto:cgs...@gm...>> wrote:

Hi Yinglong,

Thanks for the helpful example, I tried a couple more things:
1) I tried compiling amber24 in several different places, all while editing the $AMBERHOME/amber/pmemd24_src/src/pmemd/src/gti_controlVariable.i file to match what you sent. Each time I tried a run, I found the error saying the TI atom limit is 500.
2) I tried compiling amber24 without editing any files. This means the default MaxNumberTIAtom is 750 inside $AMBERHOME/amber/pmemd24_src/src/pmemd/src/gti_controlVariable.i. When I try to run GaMD-PPI with this version, my system still crashes with a TI atom limit of 500.

Am I missing something? In amber24 is there another variable that overrides the limits set inside the gti_controlVariable.i file? If so, that would explain why running GaMP-PPI shows an atom limit of 500 even though the default amber24 TI atom limit inside gti_controlVariable.i is 750. In 2) the only file I edited was the $AMBERHOME/amber/pmemd24_src/build/run_cmake file, where I edited the linux section to be

  cmake $AMBER_PREFIX/pmemd24_src \
    -DCMAKE_INSTALL_PREFIX=$AMBER_PREFIX/pmemd24 \
    -DCOMPILER=MANUAL  \
    -DMPI=TRUE -DCUDA=TRUE -DINSTALL_TESTS=TRUE \
    -DDOWNLOAD_MINICONDA=FALSE -DBUILD_PYTHON=FALSE \
    -DBUILD_PERL=FALSE -DBUILD_GUI=FALSE \
    -DPMEMD_ONLY=TRUE -DCHECK_UPDATES=FALSE \
    -DDISABLE_TOOLS=FEW -DCMAKE_C_COMPILER=gcc-12 \
    -DCMAKE_CXX_COMPILER=g++-12 -DCMAKE_Fortran_COMPILER=gfortran-12 \
    2>&1 | tee  cmake.log

to compile on the resource I am using. Let me know if there are any files I can send to help diagnose this~

Best,
Christian

On Thu, Oct 30, 2025 at 10:56 AM Miao, Yinglong <Yin...@me...<mailto:Yin...@me...>> wrote:
Hi Christian,

Sorry for the issue with running PPI-GaMD. To perhaps follow up with the post by Dr. Fabian Glaser as you mentioned in your email, I’m attaching more detailed discussion he had later on with Jinan to resolve the issue.

Hope the example “gti_controlVariable.i” and the idea of using “only the interfacial atoms of the first protein in TIMASK1/SCMASK1, and then the second FULL protein first and last atoms in bgpro2atm and edprod2atm” could help.

For “bgpro2atm and edpro2atm”, it would make sense to include ALL atoms associated with the second protein, including the PTMs.

Best,
Yinglong

Yinglong Miao, Ph.D.
Associate Professor
Department of Pharmacology and
Computational Medicine Program
University of North Carolina - Chapel Hill
Tel: 1-919-962-5696
http://miaolab.org<http://miaolab.org/>

Editor-in-Chief
npj Drug Discovery
https://www.nature.com/npjdrugdiscov

On Oct 29, 2025, at 5:03 PM, Christian Seitz <cgs...@gm...<mailto:cgs...@gm...>> wrote:

You don't often get email from cgs...@gm...<mailto:cgs...@gm...>. Learn why this is important<https://aka.ms/LearnAboutSenderIdentification>
Hello Yinglong + co,

I am trying to equilibrate a system for use in GaMD-PPI using amber24. I ran it through normal equilibration for MD (heating, minimizing, equilibration) already. I tried to start a GaMD equilibration, but immediately the system fails with a segfault:

terminate called after throwing an instance of 'std::runtime_error'
  what():  Too many TI atoms--current limit: 500 atoms

I went into $AMBERHOME/amber/pmemd24_src/src/pmemd/src/gti_controlVariable.i and changed the following parameters:

MaxNumberTIAtom = 750 -> 5250
MaxNumberTIPair = 300 -> 3000
MaxNumberREAFAtom = 500 -> 4000

I can make a fresh build of amber24 and compile with these adjusted parameters, but running it still results in a TI atom limit of 500 atoms. This issue was seen previously here<http://archive.ambermd.org/202412/0029.html>, here<https://sourceforge.net/p/gamd/mailman/gamd-discuss/thread/SA1PR03MB6450A4FB3834171AE436C154B85F2%40SA1PR03MB6450.namprd03.prod.outlook.com/#msg58743189>, here<http://archive.ambermd.org/202206/0050.html>, and here<http://archive.ambermd.org/202307/0086.html>. It appears one person<https://sourceforge.net/p/gamd/mailman/gamd-discuss/thread/36223A2E-0AC1-4D6B-8F42-E1B5C996FE4F%40uconn.edu/> was able to fix this issue in a previous version of amber, but not amber24. Do you have any ideas on what other parameters I would need to modify to increase the TI atom limit? Thank you!

I also have a question about the atom masks. When I select the atom mask bgpro2atm and edpro2atm, should this mask only include protein atoms or also PTMs, such as glycans, attached to the protein?

Thanks for your help!

Best,
Christian

--
Christian Seitz
PhD, Biochemistry & Biophysics | UCSD
Postdoc, Computer Science | UChicago
cgs...@gm...<mailto:cg...@uc...>
[www.linkedin.com/in/christianseitz21]<http://www.linkedin.com/in/christianseitz21>
_______________________________________________
GaMD-discuss mailing list
GaM...@li...<mailto:GaM...@li...>
https://lists.sourceforge.net/lists/listinfo/gamd-discuss

--
Christian Seitz
PhD, Biochemistry & Biophysics | UCSD
Postdoc, Computer Science | UChicago
cgs...@gm...<mailto:cg...@uc...>
[www.linkedin.com/in/christianseitz21]<http://www.linkedin.com/in/christianseitz21>
_______________________________________________
GaMD-discuss mailing list
GaM...@li...<mailto:GaM...@li...>
https://lists.sourceforge.net/lists/listinfo/gamd-discuss

--
Christian Seitz
PhD, Biochemistry & Biophysics | UCSD
Postdoc, Computer Science | UChicago
cgs...@gm...<mailto:cg...@uc...>
[www.linkedin.com/in/christianseitz21]<http://www.linkedin.com/in/christianseitz21>

--
Christian Seitz
PhD, Biochemistry & Biophysics | UCSD
Postdoc, Computer Science | UChicago
cgs...@gm...<mailto:cg...@uc...>
[www.linkedin.com/in/christianseitz21]<http://www.linkedin.com/in/christianseitz21>

--
Christian Seitz
PhD, Biochemistry & Biophysics | UCSD
Postdoc, Computer Science | UChicago
cgs...@gm...<mailto:cg...@uc...>

--
Christian Seitz
PhD, Biochemistry & Biophysics | UCSD
Postdoc, Computer Science | UChicago
cgs...@gm...<mailto:cg...@uc...>
[www.linkedin.com/in/christianseitz21]<http://www.linkedin.com/in/christianseitz21>
_______________________________________________
GaMD-discuss mailing list
GaM...@li...
https://lists.sourceforge.net/lists/listinfo/gamd-discuss