gamd-discuss Mailing List for GaMD
Brought to you by:
yinglongmiao
Menu
▾
▴
gamd-discuss
You can subscribe to this list here.
| 2017 |
Jan
|
Feb
|
Mar
|
Apr
|
May
|
Jun
|
Jul
|
Aug
|
Sep
(26) |
Oct
(10) |
Nov
(2) |
Dec
(5) |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 2018 |
Jan
(11) |
Feb
(7) |
Mar
(10) |
Apr
(8) |
May
(8) |
Jun
(15) |
Jul
(7) |
Aug
|
Sep
(3) |
Oct
(5) |
Nov
(8) |
Dec
|
| 2019 |
Jan
(5) |
Feb
(28) |
Mar
(14) |
Apr
(17) |
May
(7) |
Jun
(14) |
Jul
(4) |
Aug
(22) |
Sep
(13) |
Oct
(4) |
Nov
(4) |
Dec
(1) |
| 2020 |
Jan
(6) |
Feb
(3) |
Mar
(13) |
Apr
(5) |
May
(6) |
Jun
(6) |
Jul
(14) |
Aug
(3) |
Sep
(21) |
Oct
(16) |
Nov
(20) |
Dec
(10) |
| 2021 |
Jan
(14) |
Feb
(11) |
Mar
(16) |
Apr
(51) |
May
(15) |
Jun
(25) |
Jul
(8) |
Aug
(15) |
Sep
(6) |
Oct
(10) |
Nov
(7) |
Dec
(10) |
| 2022 |
Jan
(12) |
Feb
(19) |
Mar
(8) |
Apr
(18) |
May
(10) |
Jun
(16) |
Jul
(20) |
Aug
(6) |
Sep
(5) |
Oct
(13) |
Nov
(14) |
Dec
(15) |
| 2023 |
Jan
(9) |
Feb
(15) |
Mar
(5) |
Apr
(3) |
May
(38) |
Jun
(18) |
Jul
(8) |
Aug
(16) |
Sep
(21) |
Oct
(27) |
Nov
(10) |
Dec
(7) |
| 2024 |
Jan
(8) |
Feb
(19) |
Mar
(20) |
Apr
(6) |
May
(6) |
Jun
(23) |
Jul
(15) |
Aug
(11) |
Sep
(13) |
Oct
(20) |
Nov
(3) |
Dec
(8) |
| 2025 |
Jan
(3) |
Feb
(6) |
Mar
|
Apr
(10) |
May
(7) |
Jun
(13) |
Jul
(6) |
Aug
(8) |
Sep
(2) |
Oct
|
Nov
|
Dec
|
|
From: Miao, Y. <Yin...@me...> - 2025-09-30 23:04:38
|
Hi Matthew,
We discussed this in previous GaMD posts. But this was what I found with chatGPT: how to set ntave for gamd with large number of atoms?
Nice — here’s a practical, ready-to-use guide for choosing ntave for GaMD (from Yinglong Miao’s GaMD guidance + papers), tuned for large systems.
________________________________
Quick rule of thumb
Set
ntave ≈ max(50,000, 4 × N_atoms)
where N_atoms = total number of atoms in your simulation.
This recommendation comes from the Miao (GaMD) manual: ntave is the number of steps used to compute the average and standard deviation of potential energies, and it’s recommended to be roughly four times the total atom count (with 50,000 as a typical lower default).
________________________________
Why this matters (short)
* ntave controls how many simulation steps are averaged when estimating Vavg and σV (the mean and SD of the potential energy) used to compute the GaMD boost.
* Too small ntave → noisy estimates → unstable or incorrect boost parameters.
* Too large ntave → slower responsiveness to real changes in the potential statistics and requires ntcmd/nteb to be multiples of ntave. Balance is required.
________________________________
Practical examples
1. Small system (20,000 atoms)
4 × N_atoms = 80,000 ⇒ use ntave = 80,000 ( > 50k ).
2. Large system (200,000 atoms)
4 × N_atoms = 800,000 ⇒ use ntave = 800,000. For very large systems you may need ntave in the 1e6 range (some GaMD docs/tuts note defaults of 1,000,000 for certain setups).
3. 2 fs timestep context (common):
ntave = 800,000 steps × 2 fs = 1.6 ns worth of samples per average-window — reasonable for getting stable statistics for big solvated systems.
________________________________
How to set related parameters (checklist)
* Ensure ntcmd and nteb are multiples of ntave (required).
* Typical GaMD workflow (names may differ by implementation):
* ntcmdprep (equilibration prep) — allow initial relaxation.
* ntcmd (collect baseline stats) — must be a multiple of ntave.
* ntebprep (bias prep) — prep before applying bias.
* nteb (biasing stage where you update stats) — also a multiple of ntave.
* Production (nstlim) — can be set to update adaptively or run with fixed boost.
* If your Vavg/σV aren’t converging during nteb, increase nteb (and/or ntave) until they level off.
________________________________
Monitoring & validation
* Plot Vavg and σV vs. step during the preparation/biasing phase. You should see these statistics settle (flat) before full production. If they are noisy or drifting, increase ntave and/or nteb.
* Check the boost magnitude (and reweighting quality) — extreme boost values or bad reweighting indicate poor stats; if so, raise ntave or lengthen nteb.
________________________________
Extra tips for very large systems
* Start with ntave = 4 × N_atoms. If the resulting ntave is huge (>> 1e6) and you cannot afford such windows, try:
* Use multiple smaller GaMD boosts (e.g., dual-boost strategy that biases selective energies like dihedral + total or ligand-only GaMD variants), or
* Reduce the biased energy target (bias only subsystem energies when supported — e.g., LiGaMD/Pep-GaMD approaches).
These strategies are discussed in Miao’s GaMD/LiGaMD literature.
________________________________
Short template you can paste into your GaMD config
(Replace <N_atoms> and times with your values.)
# compute ntave
N_atoms=<TOTAL_ATOM_COUNT>
ntave = max(50000, 4 * N_atoms)
# ensure ntcmd and nteb are multiples of ntave
ntcmdprep = ntave
ntcmd = 10 * ntave # example: 10 windows for baseline
ntebprep = ntave
nteb = 20 * ntave # example: longer biasing prep
# production: keep boost and reweight checks in mind
________________________________
I may add that during GaMD equilibration, you can track k0P and k0D in the mdout file. If they can reach 1.0 by the end of your equilibration, that should work well for production simulations.
Hope this helps!
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
Editor-in-Chief
npj Drug Discovery
https://www.nature.com/npjdrugdiscov
On Sep 29, 2025, at 10:49 PM, Wong, Matthew via GaMD-discuss <gam...@li...> wrote:
Hi Miao Lab,
I am trying to apply GaMD to simulate a relatively large system, with 22k atoms in the protein alone and reaching 870k atoms with solvent.
I set ntave = 3000000 as it is approximately 4 times of the number of atoms, with the remaining variables all being the same (i.e. ntcmd = 3000000, nteb = 3000000, ntave = 3000000, ntcmdprep = 3000000, ntebprep = 3000000).
I’m wondering if you would suggest any adjustments to the number of steps for these processes, and if there are any general guidelines I should follow?
Thank you!
Best,
Matthew
_______________________________________________
GaMD-discuss mailing list
GaM...@li...
https://lists.sourceforge.net/lists/listinfo/gamd-discuss
|
|
From: Wong, M. <Mat...@um...> - 2025-09-30 06:00:13
|
Hi Miao Lab, I am trying to apply GaMD to simulate a relatively large system, with 22k atoms in the protein alone and reaching 870k atoms with solvent. I set ntave = 3000000 as it is approximately 4 times of the number of atoms, with the remaining variables all being the same (i.e. ntcmd = 3000000, nteb = 3000000, ntave = 3000000, ntcmdprep = 3000000, ntebprep = 3000000). I’m wondering if you would suggest any adjustments to the number of steps for these processes, and if there are any general guidelines I should follow? Thank you! Best, Matthew |
|
From: 朱峻瑶 <zhu...@st...> - 2025-08-26 12:58:39
|
Dear Prof. Miao and Dr. Victor, Thank you so much, Prof. Miao, for your clarification and for connecting me with an expert from your lab. Hello Victor, it's a pleasure to meet you, and thank you in advance for taking the time to look at my issue.I have attached the input file (job1.in), the full output file (job1.out), restart file (step7_1.rst7) and topology file (step5_input.parm7) to this email for your review. To make it easier to diagnose, I wanted to provide a full summary of my system and the problem I am facing. My system is a GPCR membrane protein, which I built using CHARMM-GUI with HMR enabled. The protein has two main domains that are non-covalently connected: a transmembrane domain and an extracellular domain. A key detail is that the extracellular domain is itself a complex of four separate proteins, which are also non-covalently bound to each other. I first performed a 500 ns conventional MD equilibration on the system, and it was stable. My goal now is to use PPI-GaMD to sample the interaction between the transmembrane and extracellular domains. The problem is that the simulation fails with a segmentation fault when I set igamd=17 for the PPI-GaMD production run. However, the simulation runs perfectly fine when I set igamd=3 for conventional GaMD using the same starting files. This is my first time attempting to use PPI-GaMD, so I am not sure what is causing this error. Any insight you could offer would be greatly appreciated.Please let me know if you have any trouble opening the attachments or if any other information would be helpful. Thank you both again for your time and willingness to help. I greatly appreciate it. Best regards, Junyao -----原始邮件----- 发件人:"Miao, Yinglong" <Yin...@me...> 发送时间:2025-08-23 21:36:13 (星期六) 收件人: 朱峻瑶 <zhu...@st...> 抄送: "Adediwura, Victor Ayo" <vad...@em...>, gamd-discuss <gam...@li...> 主题: Re: [GaMD-discuss] [PPI-GaMD Running Error] Problem running PPI-GaMD simulation - [Program received signal SIGSEGV: Segmentation fault - invalid memory reference.] Hi Junyao, In case the protein domains are non-covalently connected, PPI-GaMD would still work. I’m CC’ing Victor, a PPI-GaMD expert in our lab, and he may help you further to check the issue if you can possibly share the .in and mdout files of your simulation. 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 Editor-in-Chief npj Drug Discovery https://www.nature.com/npjdrugdiscov ======= For discussion on GaMD, please subscribe to mailing list: https://sourceforge.net/projects/gamd/lists/gamd-discuss, post emails to "gam...@li..." and help others as much as you can! ======= On Aug 22, 2025, at 9:15 AM, 朱峻瑶 <zhu...@st...> wrote: Dear Prof. Miao, Thank you for your prompt reply and helpful advice. The two domains in my simulation are indeed non-covalently connected.One of the domains is a multi-protein complex, also held together by non-covalent interactions. I am now wondering if defining a group with multiple non-covalent chains is causing the error in PPI-GaMD. Thank you for the suggestions. I will try the LiGaMD and Pep-GaMD methods you mentioned. Best regards, Junyao -----原始邮件----- 发件人: "Miao, Yinglong" <Yin...@me...> 发送时间: 2025-08-22 19:58:03 (星期五) 收件人: 朱峻瑶 <zhu...@st...> 抄送: gamd-discuss <gam...@li...> 主题: Re: [GaMD-discuss] [PPI-GaMD Running Error] Problem running PPI-GaMD simulation - [Program received signal SIGSEGV: Segmentation fault - invalid memory reference.] Dear Junyao, I was wondering whether your two protein domains are “covalently” connected. This may cause the issue for running PPI-GaMD. Note that PPI-GaMD was designed to enhance sampling “non-covalent” interactions of two separate proteins. In case you want to sample dynamic motions of a domain in a larger protein, LiGaMD3 or Pep-GaMD may help with this purpose. Best regards, 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 Editor-in-Chief npj Drug Discovery https://www.nature.com/npjdrugdiscov ======= For discussion on GaMD, please subscribe to mailing list: https://sourceforge.net/projects/gamd/lists/gamd-discuss, post emails to "gam...@li..." and help others as much as you can! ======= On Aug 20, 2025, at 5:08 AM, 朱峻瑶 <zhu...@st...> wrote: Dear GaMD Developers, My name is Junyao Zhu, and I am an undergraduate student at Dali University. I am very interested in your team's PPI-GaMD method and would like to apply it to my research. I have been following the official tutorial (GaMD-Amber: Tutorial | Miao Lab). I successfully completed the first job (conventional MD and GaMD equilibration with igamd=3). However, when I try to run the second job, the PPI-GaMD production simulation with igamd=17, the simulation fails immediately with a segmentation fault. My simulation system was built using CHARMM-GUI.The protein is a GPCR that has a transmembrane domain and a multi-subunit extracellular domain.The system is embedded in a membrane with explicit solvent.Importantly, Hydrogen Mass Repartitioning has been enabled. My goal is to use the dual-boost potential (igamd=17) to enhance the sampling between these two domains.The groups are defined as Group 1: The multi-subunit extracellular domain. Group 2: The transmembrane domain. Command used for the production run and Error Message $ pmemd.cuda -O -i job2.in -o job2.out -p step5_input.parm7 -c job1.rst7 -r job2.rst7 -x job2.nc -gamd gamd2.log Program received signal SIGSEGV: Segmentation fault - invalid memory reference. Backtrace for this error: #0 0x7c6887223960 in ??? #1 0x7c6887222ac5 in ??? #2 0x7c6886e4251f in ??? at ./signal/../sysdeps/unix/sysv/linux/x86_64/libc_sigaction.c:0 #3 0x5ac454ecb93a in ??? #4 0x5ac454d34423 in ??? #5 0x5ac454d9d543 in ??? #6 0x5ac454d7a514 in ??? #7 0x5ac454cb015e in ??? #8 0x7c6886e29d8f in __libc_start_call_main at ../sysdeps/nptl/libc_start_call_main.h:58 #9 0x7c6886e29e3f in __libc_start_main_impl at ../csu/libc-start.c:392 #10 0x5ac454ccc044 in ??? #11 0xffffffffffffffff in ??? Segmentation fault (core dumped) Here are the GaMD parameters from my input file, ! PPI-GaMD Prod input &cntrl nstlim = 125000000, dt=0.004, irest=1, ntx=5, ntb=2, ntp=3, barostat=2, pres0=1.0, csurften=3, gamma_ten=0.0, ninterface=2 ntt=3, gamma_ln=1.0, temp0=310.0, ntc=2, ntf=1, cut=12.0, ig=-1, ntpr=100, ntwx=100, ntwr=5000, igamd=17, iE=1, irest_gamd=1, ntcmd=0, ntcmdprep=0, nteb=0, ntebprep=0, ntave=100000, sigma0P=6.0, sigma0D=6.0, timask1=':1-596', scmask1=':1-596', timask2=':597-947', scmask2=':597-947', &end I suspect the issue might be related to the complexity of the GPCR membrane system, the atom mask definitions, or a specific interaction with the HMR settings, but I have been unable to solve it. Any guidance you could offer would be immensely helpful. Please do not hesitate to contact me if you require any additional files or information. Thank you for your time. Sincerely, Junyao Zhu _______________________________________________ GaMD-discuss mailing list GaM...@li... https://lists.sourceforge.net/lists/listinfo/gamd-discuss |
|
From: Miao, Y. <Yin...@me...> - 2025-08-23 16:09:26
|
Hi Junyao, In case the protein domains are non-covalently connected, PPI-GaMD would still work. I’m CC’ing Victor, a PPI-GaMD expert in our lab, and he may help you further to check the issue if you can possibly share the .in and mdout files of your simulation. 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 Editor-in-Chief npj Drug Discovery https://www.nature.com/npjdrugdiscov ======= For discussion on GaMD, please subscribe to mailing list: https://sourceforge.net/projects/gamd/lists/gamd-discuss, post emails to "gam...@li..." and help others as much as you can! ======= On Aug 22, 2025, at 9:15 AM, 朱峻瑶 <zhu...@st...> wrote: Dear Prof. Miao, Thank you for your prompt reply and helpful advice. The two domains in my simulation are indeed non-covalently connected.One of the domains is a multi-protein complex, also held together by non-covalent interactions. I am now wondering if defining a group with multiple non-covalent chains is causing the error in PPI-GaMD. Thank you for the suggestions. I will try the LiGaMD and Pep-GaMD methods you mentioned. Best regards, Junyao -----原始邮件----- 发件人: "Miao, Yinglong" <Yin...@me...> 发送时间: 2025-08-22 19:58:03 (星期五) 收件人: 朱峻瑶 <zhu...@st...> 抄送: gamd-discuss <gam...@li...> 主题: Re: [GaMD-discuss] [PPI-GaMD Running Error] Problem running PPI-GaMD simulation - [Program received signal SIGSEGV: Segmentation fault - invalid memory reference.] Dear Junyao, I was wondering whether your two protein domains are “covalently” connected. This may cause the issue for running PPI-GaMD. Note that PPI-GaMD was designed to enhance sampling “non-covalent” interactions of two separate proteins. In case you want to sample dynamic motions of a domain in a larger protein, LiGaMD3 or Pep-GaMD may help with this purpose. Best regards, 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 Editor-in-Chief npj Drug Discovery https://www.nature.com/npjdrugdiscov ======= For discussion on GaMD, please subscribe to mailing list: https://sourceforge.net/projects/gamd/lists/gamd-discuss, post emails to "gam...@li..." and help others as much as you can! ======= On Aug 20, 2025, at 5:08 AM, 朱峻瑶 <zhu...@st...> wrote: Dear GaMD Developers, My name is Junyao Zhu, and I am an undergraduate student at Dali University. I am very interested in your team's PPI-GaMD method and would like to apply it to my research. I have been following the official tutorial (GaMD-Amber: Tutorial | Miao Lab). I successfully completed the first job (conventional MD and GaMD equilibration with igamd=3). However, when I try to run the second job, the PPI-GaMD production simulation with igamd=17, the simulation fails immediately with a segmentation fault. My simulation system was built using CHARMM-GUI.The protein is a GPCR that has a transmembrane domain and a multi-subunit extracellular domain.The system is embedded in a membrane with explicit solvent.Importantly, Hydrogen Mass Repartitioning has been enabled. My goal is to use the dual-boost potential (igamd=17) to enhance the sampling between these two domains.The groups are defined as Group 1: The multi-subunit extracellular domain. Group 2: The transmembrane domain. Command used for the production run and Error Message $ pmemd.cuda -O -i job2.in -o job2.out -p step5_input.parm7 -c job1.rst7 -r job2.rst7 -x job2.nc -gamd gamd2.log Program received signal SIGSEGV: Segmentation fault - invalid memory reference. Backtrace for this error: #0 0x7c6887223960 in ??? #1 0x7c6887222ac5 in ??? #2 0x7c6886e4251f in ??? at ./signal/../sysdeps/unix/sysv/linux/x86_64/libc_sigaction.c:0 #3 0x5ac454ecb93a in ??? #4 0x5ac454d34423 in ??? #5 0x5ac454d9d543 in ??? #6 0x5ac454d7a514 in ??? #7 0x5ac454cb015e in ??? #8 0x7c6886e29d8f in __libc_start_call_main at ../sysdeps/nptl/libc_start_call_main.h:58 #9 0x7c6886e29e3f in __libc_start_main_impl at ../csu/libc-start.c:392 #10 0x5ac454ccc044 in ??? #11 0xffffffffffffffff in ??? Segmentation fault (core dumped) Here are the GaMD parameters from my input file, ! PPI-GaMD Prod input &cntrl nstlim = 125000000, dt=0.004, irest=1, ntx=5, ntb=2, ntp=3, barostat=2, pres0=1.0, csurften=3, gamma_ten=0.0, ninterface=2 ntt=3, gamma_ln=1.0, temp0=310.0, ntc=2, ntf=1, cut=12.0, ig=-1, ntpr=100, ntwx=100, ntwr=5000, igamd=17, iE=1, irest_gamd=1, ntcmd=0, ntcmdprep=0, nteb=0, ntebprep=0, ntave=100000, sigma0P=6.0, sigma0D=6.0, timask1=':1-596', scmask1=':1-596', timask2=':597-947', scmask2=':597-947', &end I suspect the issue might be related to the complexity of the GPCR membrane system, the atom mask definitions, or a specific interaction with the HMR settings, but I have been unable to solve it. Any guidance you could offer would be immensely helpful. Please do not hesitate to contact me if you require any additional files or information. Thank you for your time. Sincerely, Junyao Zhu _______________________________________________ GaMD-discuss mailing list GaM...@li...<mailto:GaM...@li...> https://lists.sourceforge.net/lists/listinfo/gamd-discuss |
|
From: Miao, Y. <Yin...@me...> - 2025-08-23 13:32:52
|
Hi Lei, I’m CC’ing Keya, a GaMD expert in our lab, and she may help you to double check the issue? 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 Editor-in-Chief npj Drug Discovery https://www.nature.com/npjdrugdiscov ======= For discussion on GaMD, please subscribe to mailing list: https://sourceforge.net/projects/gamd/lists/gamd-discuss, post emails to "gam...@li..." and help others as much as you can! ======= On Aug 22, 2025, at 12:47 PM, 鲍磊 <bo...@wh...> wrote: Dear Prof. Miao, I have attached the coordinate and force field files for the MgCl₂ solution system simulation, along with the input and output files for both cMD and GaMD. Could you please help me analyze the cause of the issue? Thank you very much for your guidance. Best, Lei -----原始邮件----- 发件人: "Miao, Yinglong" <Yin...@me...> 发送时间: 2025-08-22 20:04:01 (星期五) 收件人: 鲍磊 <bo...@wh...> 抄送: gamd-discuss <gam...@li...> 主题: Re: Some questions about GaMD Dear Lei, Could it be that GaMD imposes specific requirements on the simulation box shape? (I have been using a truncated octahedral box throughout.) No, GaMD shouldn’t affect anything in the cMD stage, including the simulation box. Only potential statistics are collected during cMD. If you can share a copy of the simulation mdout file, we can probably look into it for more detail. 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 Editor-in-Chief npj Drug Discovery https://www.nature.com/npjdrugdiscov ======= For discussion on GaMD, please subscribe to mailing list: https://sourceforge.net/projects/gamd/lists/gamd-discuss, post emails to "gam...@li..." and help others as much as you can! ======= On Aug 21, 2025, at 4:09 AM, 鲍磊 <bo...@wh...> wrote: Dear Prof. Miao, Thank you very much for taking the time to reply amidst your busy schedule. To further investigate the issue, I conducted additional simulations on a pure magnesium chloride solution system without any biomolecules. In these simulations, I did not apply the Hydrogen Mass Repartitioning (HMR) method and maintained a 2 fs integration time step. However, once GaMD was enabled (with igamd=3), the system density became higher than that in conventional MD simulations, and the exchange rate of water molecules in the hydration shell of magnesium ions increased significantly—consistent with the observations in the earlier protein simulations. This has left me quite perplexed, as I am unsure where the problem might lie. In the simulation parameters, I only enabled GaMD while keeping all other settings identical. Yet, even during the conventional equilibration phase, noticeable differences from CMD emerged. Could it be that GaMD imposes specific requirements on the simulation box shape? (I have been using a truncated octahedral box throughout.) Since studying protein aggregation is highly time-consuming, I had hoped to use GaMD to accelerate the process while ensuring the physical environment remains consistent, allowing for a fair comparison with CMD results. However, the current issue significantly impacts the outcomes, and I have been struggling to resolve it. I would be immensely grateful if you could offer any additional suggestions or insights. Thank you once again for your guidance and support. Best regards, Lei -----原始邮件----- 发件人: "Miao, Yinglong" <Yin...@me...<mailto:Yin...@me...>> 发送时间: 2025-08-20 09:57:55 (星期三) 收件人: 鲍磊 <bo...@wh...<mailto:bo...@wh...>> 抄送: gamd-discuss <gam...@li...<mailto:gam...@li...>> 主题: Re: Some questions about GaMD Dear Lei, Thanks for your interest in using GaMD. Among the system settings you mentioned, I don’t see any major issue. For your questions: Does GaMD inadvertently introduce acceleration effects even during the ntcmd phase? No. There should be no boost applied in ntcmd steps. Could dihedral potential boosting, intended for proteins, also affect rigid molecules like magnesium ions and water? No. As the name indicates, the "dihedral boost” (igamd=2) adds boost potential to dihedrals in the system, which are not found in the ions or water. If neither of the above is the case, might there be an issue with my simulation setup? I noticed you used dt=0.004, for which you may have used hydrogen mass repartitioning (HMR) in your system. This has not been tested sufficiently with GaMD. It might help if you can start with dt=0.002 in GaMD simulations. You may also try igamd=1 (total boost) and igamd=3 (dual-boost), especially the latter, and see how they work for you. Hope this helps, 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<https://mailgw.whu.edu.cn/linkserver?dest=aHR0cDovL21pYW9sYWIub3JnLw%3D%3D%0D%0A&tid=_____wBXEyrNXKhoNvDcCA--.5279S2&rcpt=bolly%40whu.edu.cn&ifnotice=1&rindex=0> Editor-in-Chief npj Drug Discovery https://www.nature.com/npjdrugdiscov On Aug 18, 2025, at 3:46 AM, 鲍磊 <bo...@wh...<mailto:bo...@wh...>> wrote: [You don't often get email from bo...@wh...<mailto:bo...@wh...>. Learn why this is important at https://aka.ms/LearnAboutSenderIdentification ] Dear Professor Miao, I hope this message finds you well. Please forgive me for disturbing you, but I recently encountered some perplexing issues while applying the GaMD method you developed in my simulations using AMBER20, and I would greatly appreciate your insights. The system I am studying is relatively straightforward: it involves the aggregation of 10 intrinsically disordered peptide chains in a magnesium chloride solution. For the protein force field, I chose a combination of ff14sb and ff14IDPSFF to account for the disordered nature of the peptides. For magnesium ions, I employed the Li-Merz 12-6-4 force field to describe their polarization effects, paired with the SPC/E water model. In conventional molecular dynamics (MD) simulations, all results appear normal and align well with experimental data. However, when I attempted to use GaMD to accelerate the simulation and explore the protein aggregation process over longer timescales, some unexpected issues arose. To isolate the problem, I controlled all variables and compared the results between conventional MD (CMD) and GaMD. Specifically, when I set igamd=2 (i.e., enabling dihedral potential boosting to focus solely on accelerating protein conformational changes without affecting other molecules in the system), I noticed discrepancies even during the initial ntcmd steps. Although the output file correctly indicated igamd=0 during this equilibration phase, the results still deviated significantly from those of a pure igamd=0 (CMD) simulation. For instance, during the ntcmd phase of GaMD, the simulation box exhibited smaller dimensions, leading to higher system density (see attached figures for details). Further analysis revealed that the exchange rate of water molecules in the hydration shell around magnesium ions also changed markedly during the GaMD ntcmd phase, exhibiting a higher exchange rate that deviated from experimental values, whereas the conventional MD results remained consistent. This has led me to the following questions: Does GaMD inadvertently introduce acceleration effects even during the ntcmd phase? Could dihedral potential boosting, intended for proteins, also affect rigid molecules like magnesium ions and water? If neither of the above is the case, might there be an issue with my simulation setup? I am deeply puzzled by these observations and would be immensely grateful for your expert guidance. Thank you once again for your time and consideration. Best regards, Sincerely, Lei Bao [cid:4e2...@na...] [cid:87d...@na...] [cid:db4...@na...] [cid:bff...@na...] [cid:960...@na...] [cid:258...@na...] [cid:fb9...@na...] [cid:598...@na...] |
|
From: Miao, Y. <Yin...@me...> - 2025-08-22 17:34:24
|
Dear Junyao, I was wondering whether your two protein domains are “covalently” connected. This may cause the issue for running PPI-GaMD. Note that PPI-GaMD was designed to enhance sampling “non-covalent” interactions of two separate proteins. In case you want to sample dynamic motions of a domain in a larger protein, LiGaMD3 or Pep-GaMD may help with this purpose. Best regards, 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 Editor-in-Chief npj Drug Discovery https://www.nature.com/npjdrugdiscov ======= For discussion on GaMD, please subscribe to mailing list: https://sourceforge.net/projects/gamd/lists/gamd-discuss, post emails to "gam...@li..." and help others as much as you can! ======= On Aug 20, 2025, at 5:08 AM, 朱峻瑶 <zhu...@st...> wrote: Dear GaMD Developers, My name is Junyao Zhu, and I am an undergraduate student at Dali University. I am very interested in your team's PPI-GaMD method and would like to apply it to my research. I have been following the official tutorial (GaMD-Amber: Tutorial | Miao Lab). I successfully completed the first job (conventional MD and GaMD equilibration with igamd=3). However, when I try to run the second job, the PPI-GaMD production simulation with igamd=17, the simulation fails immediately with a segmentation fault. My simulation system was built using CHARMM-GUI.The protein is a GPCR that has a transmembrane domain and a multi-subunit extracellular domain.The system is embedded in a membrane with explicit solvent.Importantly, Hydrogen Mass Repartitioning has been enabled. My goal is to use the dual-boost potential (igamd=17) to enhance the sampling between these two domains.The groups are defined as Group 1: The multi-subunit extracellular domain. Group 2: The transmembrane domain. Command used for the production run and Error Message $ pmemd.cuda -O -i job2.in -o job2.out -p step5_input.parm7 -c job1.rst7 -r job2.rst7 -x job2.nc -gamd gamd2.log Program received signal SIGSEGV: Segmentation fault - invalid memory reference. Backtrace for this error: #0 0x7c6887223960 in ??? #1 0x7c6887222ac5 in ??? #2 0x7c6886e4251f in ??? at ./signal/../sysdeps/unix/sysv/linux/x86_64/libc_sigaction.c:0 #3 0x5ac454ecb93a in ??? #4 0x5ac454d34423 in ??? #5 0x5ac454d9d543 in ??? #6 0x5ac454d7a514 in ??? #7 0x5ac454cb015e in ??? #8 0x7c6886e29d8f in __libc_start_call_main at ../sysdeps/nptl/libc_start_call_main.h:58 #9 0x7c6886e29e3f in __libc_start_main_impl at ../csu/libc-start.c:392 #10 0x5ac454ccc044 in ??? #11 0xffffffffffffffff in ??? Segmentation fault (core dumped) Here are the GaMD parameters from my input file, ! PPI-GaMD Prod input &cntrl nstlim = 125000000, dt=0.004, irest=1, ntx=5, ntb=2, ntp=3, barostat=2, pres0=1.0, csurften=3, gamma_ten=0.0, ninterface=2 ntt=3, gamma_ln=1.0, temp0=310.0, ntc=2, ntf=1, cut=12.0, ig=-1, ntpr=100, ntwx=100, ntwr=5000, igamd=17, iE=1, irest_gamd=1, ntcmd=0, ntcmdprep=0, nteb=0, ntebprep=0, ntave=100000, sigma0P=6.0, sigma0D=6.0, timask1=':1-596', scmask1=':1-596', timask2=':597-947', scmask2=':597-947', &end I suspect the issue might be related to the complexity of the GPCR membrane system, the atom mask definitions, or a specific interaction with the HMR settings, but I have been unable to solve it. Any guidance you could offer would be immensely helpful. Please do not hesitate to contact me if you require any additional files or information. Thank you for your time. Sincerely, Junyao Zhu _______________________________________________ GaMD-discuss mailing list GaM...@li...<mailto:GaM...@li...> https://lists.sourceforge.net/lists/listinfo/gamd-discuss |
|
From: Miao, Y. <Yin...@me...> - 2025-08-22 13:37:09
|
Dear Lei, Could it be that GaMD imposes specific requirements on the simulation box shape? (I have been using a truncated octahedral box throughout.) No, GaMD shouldn’t affect anything in the cMD stage, including the simulation box. Only potential statistics are collected during cMD. If you can share a copy of the simulation mdout file, we can probably look into it for more detail. 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 Editor-in-Chief npj Drug Discovery https://www.nature.com/npjdrugdiscov ======= For discussion on GaMD, please subscribe to mailing list: https://sourceforge.net/projects/gamd/lists/gamd-discuss, post emails to "gam...@li..." and help others as much as you can! ======= On Aug 21, 2025, at 4:09 AM, 鲍磊 <bo...@wh...> wrote: Dear Prof. Miao, Thank you very much for taking the time to reply amidst your busy schedule. To further investigate the issue, I conducted additional simulations on a pure magnesium chloride solution system without any biomolecules. In these simulations, I did not apply the Hydrogen Mass Repartitioning (HMR) method and maintained a 2 fs integration time step. However, once GaMD was enabled (with igamd=3), the system density became higher than that in conventional MD simulations, and the exchange rate of water molecules in the hydration shell of magnesium ions increased significantly—consistent with the observations in the earlier protein simulations. This has left me quite perplexed, as I am unsure where the problem might lie. In the simulation parameters, I only enabled GaMD while keeping all other settings identical. Yet, even during the conventional equilibration phase, noticeable differences from CMD emerged. Could it be that GaMD imposes specific requirements on the simulation box shape? (I have been using a truncated octahedral box throughout.) Since studying protein aggregation is highly time-consuming, I had hoped to use GaMD to accelerate the process while ensuring the physical environment remains consistent, allowing for a fair comparison with CMD results. However, the current issue significantly impacts the outcomes, and I have been struggling to resolve it. I would be immensely grateful if you could offer any additional suggestions or insights. Thank you once again for your guidance and support. Best regards, Lei -----原始邮件----- 发件人: "Miao, Yinglong" <Yin...@me...<mailto:Yin...@me...>> 发送时间: 2025-08-20 09:57:55 (星期三) 收件人: 鲍磊 <bo...@wh...<mailto:bo...@wh...>> 抄送: gamd-discuss <gam...@li...<mailto:gam...@li...>> 主题: Re: Some questions about GaMD Dear Lei, Thanks for your interest in using GaMD. Among the system settings you mentioned, I don’t see any major issue. For your questions: Does GaMD inadvertently introduce acceleration effects even during the ntcmd phase? No. There should be no boost applied in ntcmd steps. Could dihedral potential boosting, intended for proteins, also affect rigid molecules like magnesium ions and water? No. As the name indicates, the "dihedral boost” (igamd=2) adds boost potential to dihedrals in the system, which are not found in the ions or water. If neither of the above is the case, might there be an issue with my simulation setup? I noticed you used dt=0.004, for which you may have used hydrogen mass repartitioning (HMR) in your system. This has not been tested sufficiently with GaMD. It might help if you can start with dt=0.002 in GaMD simulations. You may also try igamd=1 (total boost) and igamd=3 (dual-boost), especially the latter, and see how they work for you. Hope this helps, 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 Aug 18, 2025, at 3:46 AM, 鲍磊 <bo...@wh...<mailto:bo...@wh...>> wrote: [You don't often get email from bo...@wh...<mailto:bo...@wh...>. Learn why this is important at https://aka.ms/LearnAboutSenderIdentification ] Dear Professor Miao, I hope this message finds you well. Please forgive me for disturbing you, but I recently encountered some perplexing issues while applying the GaMD method you developed in my simulations using AMBER20, and I would greatly appreciate your insights. The system I am studying is relatively straightforward: it involves the aggregation of 10 intrinsically disordered peptide chains in a magnesium chloride solution. For the protein force field, I chose a combination of ff14sb and ff14IDPSFF to account for the disordered nature of the peptides. For magnesium ions, I employed the Li-Merz 12-6-4 force field to describe their polarization effects, paired with the SPC/E water model. In conventional molecular dynamics (MD) simulations, all results appear normal and align well with experimental data. However, when I attempted to use GaMD to accelerate the simulation and explore the protein aggregation process over longer timescales, some unexpected issues arose. To isolate the problem, I controlled all variables and compared the results between conventional MD (CMD) and GaMD. Specifically, when I set igamd=2 (i.e., enabling dihedral potential boosting to focus solely on accelerating protein conformational changes without affecting other molecules in the system), I noticed discrepancies even during the initial ntcmd steps. Although the output file correctly indicated igamd=0 during this equilibration phase, the results still deviated significantly from those of a pure igamd=0 (CMD) simulation. For instance, during the ntcmd phase of GaMD, the simulation box exhibited smaller dimensions, leading to higher system density (see attached figures for details). Further analysis revealed that the exchange rate of water molecules in the hydration shell around magnesium ions also changed markedly during the GaMD ntcmd phase, exhibiting a higher exchange rate that deviated from experimental values, whereas the conventional MD results remained consistent. This has led me to the following questions: Does GaMD inadvertently introduce acceleration effects even during the ntcmd phase? Could dihedral potential boosting, intended for proteins, also affect rigid molecules like magnesium ions and water? If neither of the above is the case, might there be an issue with my simulation setup? I am deeply puzzled by these observations and would be immensely grateful for your expert guidance. Thank you once again for your time and consideration. Best regards, Sincerely, Lei Bao [gamd-dihedral-in.png][gamd-dihedral-out.png][pure-cmd-in.png][pure-cmd-out.png] [mgcl2-cmd-in.png][mgcl2-cmd-out.png][mgcl2-gamd-in.png][mgcl2-gamd-out.png] |
|
From: Christine Q. <chr...@uc...> - 2025-08-21 20:35:54
|
Dear GaMD Developers, I hope you are doing well. I was wondering if there are any recommended configurations or practices that could help speed up GaMD jobs. I am running a protein–membrane system with slightly over 1 million atoms. Because I am working on a shared cluster with a strict wallclock limit, the GaMD equilibration step cannot finish within a single job. I understand that the equilibration is recommended to be completed in one continuous run rather than restarting, so I was wondering if you might have any suggestions for handling this situation. Are there specific configuration settings (# of CPUs, GPUs) that could help speed up the run? Thank you very much for your time and help! All the best, Christine Qian |
|
From: 朱峻瑶 <zhu...@st...> - 2025-08-20 09:09:22
|
Dear GaMD Developers, My name is Junyao Zhu, and I am an undergraduate student at Dali University. I am very interested in your team's PPI-GaMD method and would like to apply it to my research. I have been following the official tutorial (GaMD-Amber: Tutorial | Miao Lab). I successfully completed the first job (conventional MD and GaMD equilibration with igamd=3). However, when I try to run the second job, the PPI-GaMD production simulation with igamd=17, the simulation fails immediately with a segmentation fault. My simulation system was built using CHARMM-GUI.The protein is a GPCR that has a transmembrane domain and a multi-subunit extracellular domain.The system is embedded in a membrane with explicit solvent.Importantly, Hydrogen Mass Repartitioning has been enabled. My goal is to use the dual-boost potential (igamd=17) to enhance the sampling between these two domains.The groups are defined as Group 1: The multi-subunit extracellular domain. Group 2: The transmembrane domain. Command used for the production run and Error Message $ pmemd.cuda -O -i job2.in -o job2.out -p step5_input.parm7 -c job1.rst7 -r job2.rst7 -x job2.nc -gamd gamd2.log Program received signal SIGSEGV: Segmentation fault - invalid memory reference. Backtrace for this error: #0 0x7c6887223960 in ??? #1 0x7c6887222ac5 in ??? #2 0x7c6886e4251f in ??? at ./signal/../sysdeps/unix/sysv/linux/x86_64/libc_sigaction.c:0 #3 0x5ac454ecb93a in ??? #4 0x5ac454d34423 in ??? #5 0x5ac454d9d543 in ??? #6 0x5ac454d7a514 in ??? #7 0x5ac454cb015e in ??? #8 0x7c6886e29d8f in __libc_start_call_main at ../sysdeps/nptl/libc_start_call_main.h:58 #9 0x7c6886e29e3f in __libc_start_main_impl at ../csu/libc-start.c:392 #10 0x5ac454ccc044 in ??? #11 0xffffffffffffffff in ??? Segmentation fault (core dumped) Here are the GaMD parameters from my input file, ! PPI-GaMD Prod input &cntrl nstlim = 125000000, dt=0.004, irest=1, ntx=5, ntb=2, ntp=3, barostat=2, pres0=1.0, csurften=3, gamma_ten=0.0, ninterface=2 ntt=3, gamma_ln=1.0, temp0=310.0, ntc=2, ntf=1, cut=12.0, ig=-1, ntpr=100, ntwx=100, ntwr=5000, igamd=17, iE=1, irest_gamd=1, ntcmd=0, ntcmdprep=0, nteb=0, ntebprep=0, ntave=100000, sigma0P=6.0, sigma0D=6.0, timask1=':1-596', scmask1=':1-596', timask2=':597-947', scmask2=':597-947', &end I suspect the issue might be related to the complexity of the GPCR membrane system, the atom mask definitions, or a specific interaction with the HMR settings, but I have been unable to solve it. Any guidance you could offer would be immensely helpful. Please do not hesitate to contact me if you require any additional files or information. Thank you for your time. Sincerely, Junyao Zhu |
|
From: Miao, Y. <Yin...@me...> - 2025-08-20 02:32:01
|
Dear Lei, Thanks for your interest in using GaMD. Among the system settings you mentioned, I don’t see any major issue. For your questions: Does GaMD inadvertently introduce acceleration effects even during the ntcmd phase? No. There should be no boost applied in ntcmd steps. Could dihedral potential boosting, intended for proteins, also affect rigid molecules like magnesium ions and water? No. As the name indicates, the "dihedral boost” (igamd=2) adds boost potential to dihedrals in the system, which are not found in the ions or water. If neither of the above is the case, might there be an issue with my simulation setup? I noticed you used dt=0.004, for which you may have used hydrogen mass repartitioning (HMR) in your system. This has not been tested sufficiently with GaMD. It might help if you can start with dt=0.002 in GaMD simulations. You may also try igamd=1 (total boost) and igamd=3 (dual-boost), especially the latter, and see how they work for you. Hope this helps, 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 Editor-in-Chief npj Drug Discovery https://www.nature.com/npjdrugdiscov On Aug 18, 2025, at 3:46 AM, 鲍磊 <bo...@wh...> wrote: [You don't often get email from bo...@wh.... Learn why this is important at https://aka.ms/LearnAboutSenderIdentification ] Dear Professor Miao, I hope this message finds you well. Please forgive me for disturbing you, but I recently encountered some perplexing issues while applying the GaMD method you developed in my simulations using AMBER20, and I would greatly appreciate your insights. The system I am studying is relatively straightforward: it involves the aggregation of 10 intrinsically disordered peptide chains in a magnesium chloride solution. For the protein force field, I chose a combination of ff14sb and ff14IDPSFF to account for the disordered nature of the peptides. For magnesium ions, I employed the Li-Merz 12-6-4 force field to describe their polarization effects, paired with the SPC/E water model. In conventional molecular dynamics (MD) simulations, all results appear normal and align well with experimental data. However, when I attempted to use GaMD to accelerate the simulation and explore the protein aggregation process over longer timescales, some unexpected issues arose. To isolate the problem, I controlled all variables and compared the results between conventional MD (CMD) and GaMD. Specifically, when I set igamd=2 (i.e., enabling dihedral potential boosting to focus solely on accelerating protein conformational changes without affecting other molecules in the system), I noticed discrepancies even during the initial ntcmd steps. Although the output file correctly indicated igamd=0 during this equilibration phase, the results still deviated significantly from those of a pure igamd=0 (CMD) simulation. For instance, during the ntcmd phase of GaMD, the simulation box exhibited smaller dimensions, leading to higher system density (see attached figures for details). Further analysis revealed that the exchange rate of water molecules in the hydration shell around magnesium ions also changed markedly during the GaMD ntcmd phase, exhibiting a higher exchange rate that deviated from experimental values, whereas the conventional MD results remained consistent. This has led me to the following questions: Does GaMD inadvertently introduce acceleration effects even during the ntcmd phase? Could dihedral potential boosting, intended for proteins, also affect rigid molecules like magnesium ions and water? If neither of the above is the case, might there be an issue with my simulation setup? I am deeply puzzled by these observations and would be immensely grateful for your expert guidance. Thank you once again for your time and consideration. Best regards, Sincerely, Lei Bao [gamd-dihedral-in.png][gamd-dihedral-out.png][pure-cmd-in.png][pure-cmd-out.png] |
|
From: Li, J. <JO...@pi...> - 2025-07-21 15:18:01
|
Hi Fu, Thank you so much!!! The "timask1" and "scmask1" are where I did wrong and after I corrected them, I can see the ligand unbound. Thank you again and hope you have a wonderful week, Qianwei ________________________________ From: Xiao, Fu <Fu...@me...> Sent: Friday, July 18, 2025 6:51 PM To: Li, Joyce <JO...@pi...> Cc: Miao, Yinglong <Yin...@me...>; Joshi, Keya <ke...@em...>; gam...@li... <gam...@li...> Subject: Re: [GaMD-discuss] Question about proper sigma0P value You don't often get email from fu...@me.... Learn why this is important<https://aka.ms/LearnAboutSenderIdentification> Hi Qianwei, I think the parameters of "timask1" and "scmask1" in your input file are a bit strange. These two parameters are the sequence numbers of the ligand in the amber format protein complex (pdb4amber command line can do this). Similar to the red box marked in the following picture (BEN is a small molecule). [cid:bc75a470-c323-4303-88c7-ed6455b0cb77] Small molecules are represented by an integer number, while peptides/antibodies are indicated by a sequence number such as xxx-xxx. In addition, iE = 2 can be deleted. Maybe you can try to check your parameters carefully. Best, Fu Xiao ________________________________ From: Miao, Yinglong <Yin...@me...> Sent: Friday, July 18, 2025 17:05 To: Li, Joyce <JO...@pi...> Cc: gamd-discuss <gam...@li...>; Xiao, Fu <Fu...@me...>; Joshi, Keya <ke...@em...> Subject: Re: [GaMD-discuss] Question about proper sigma0P value Hi Qianwei, I’m CC’ing Fu and Keya, experts in LiGaMD3 in our lab, and they may be able to help you more. 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 Jul 17, 2025, at 6:36 PM, Li, Joyce via GaMD-discuss <gam...@li...> wrote: Hello, I am trying to use LiGAMD3 for measuring the binding kinetics between protein and ligand. The system I use is PDB 5JID. The original system has two ligands, but I kept only one for the GAMD. I followed the discussions in paper (10.1021/acs.jctc.4c00502) and tried to find the right sigma0p for unbinding the ligand. But even if I set the sigma0P to 8, the ligand still binds with the protein and the ligand RMSD is lower than 2 A after 100 ns production run. Attached are the codes for the simulation (pasted below as well) and the log file for the 100 ns production when P is 8. Did I do something wrong, or should I go for an even much higher P value? I have been trying this for months but not seenthe ligand to move. So I will appreciate any suggestions! Thank you, Best, Qianwei equbrilium cMD and GaMD equilbrium &cntrl ig = -1, ! Pseudo-random number generator (random seed based on time) ntx = 1, ! no initial coordinates and velocity(ASCII) imin = 0, ! Minimization (No) irest = 0, ! Restart the Simulation? (no) use previous velocity cut = 8, ! Non-bonding Cut-off (8 A) ntc = 2, ! SHAKE bond length constraints (constrain bonds with H) ntf = 1, ! compelete interaction is calculated tol = 0.000001, !relatve geometrical tolerance in shake dt = 0.002, ! timestep = 2fs iwrap = 1, !set for long runs to keep coordinate from overflowing nstlim = 44000000, ! Simulation length (sum of ntcmd and nteb total 88 ns) ntb = 2, ! PBC (Constant pressure) ntp = 1, ! isotropic position scaling barostat = 1, ! for control pressure, Berendsen ntt = 3, ! Temperature scaling (Stochastic version of Berendsen thermostat) gamma_ln=5.0, temp0 = 300, ! Reference temperature (300K, physiological temperature) tempi = 300.0, !Initial temperature ntpr = 500000, ! Energy save interval (every 500000 steps or 1 ns) ntwx = 500000, ! Coordinate/trajectory save interval (every 500000 steps or 1 ns) ntwprt = 7101, !number of atoms to be saved ntwr = 5000000, ! Write restart file every 10 ns igamd = 28, iE = 2, !thershold energy to the lower bound irest_gamd = 0, !not a restart gamd ntcmd = 4000000, !8 ns cMD nteb = 40000000, !80 ns GAMD ntave = 400000, !four times total number of atoms ntcmdprep = 800000, !1.6 ns number of cMD not collected for statistics calculation ntebprep = 800000, !1.6 ns GAMD for preperation sigma0P = 8.0, sigma0D = 6.0, sigma0B = 6.0, iEP=2, iED=2, iEB=1, bgpro2atm=1, edpro2atm=7076, gti_cpu_output=0, gti_add_sc=1, icfe = 1, ifsc = 1, timask1=':7077-7101', scmask1=':7077-7101', timask2='', scmask2='', &end END production GaMD production &cntrl ig = -1, ! Pseudo-random number generator (random seed based on time) ntx = 1, ! no initial coordinates and velocity(ASCII) imin = 0, ! Minimization (No) irest = 0, ! Restart the Simulation? (no) use previous velocity cut = 8, ! Non-bonding Cut-off (8 A) ntc = 2, ! SHAKE bond length constraints (constrain bonds with H) ntf = 1, ! compelete interaction is calculated tol = 0.000001, !relatve geometrical tolerance in shake dt = 0.002, ! timestep = 2fs iwrap = 1, !set for long runs to keep coordinate from overflowing nstlim = 50000000, ! Simulation length (sum of ntcmd and nteb total 100 ns) ntb = 2, ! PBC (Constant pressure) ntp = 1, ! isotropic position scaling barostat = 1, ! for control pressure, Berendsen ntt = 3, ! Temperature scaling (Stochastic version of Berendsen thermostat) gamma_ln=5.0, temp0 = 300, ! Reference temperature (300K, physiological temperature) tempi = 300.0, !Initial temperature ntpr = 250000, ! Energy save interval (every 250000 steps or 500ps) ntwx = 500, ! Coordinate/trajectory save interval (every 500 steps or 1ps) ntwprt = 7101, !number of atoms to be saved ntwr = 5000000, ! Write restart file every 10 ns igamd = 28, iE = 2, !thershold energy to the lower bound irest_gamd = 1, !a restart gamd ntcmd = 0, !no cMD nteb = 0, !0 GAMD ntave = 400000, !four times total number of atoms ntcmdprep = 0, !no cMD preperation ntebprep = 0, !noGAMD for preperation sigma0P = 8.0, sigma0D = 6.0, sigma0B = 6.0, iEP=2, iED=2, iEB=1, bgpro2atm=1, edpro2atm=7076, gti_cpu_output=0, gti_add_sc=1, icfe = 1, ifsc = 1, timask1=':7077-7101', scmask1=':7077-7101', timask2='', scmask2='', &end END _______________________________________________ GaMD-discuss mailing list GaM...@li...<mailto:GaM...@li...> https://lists.sourceforge.net/lists/listinfo/gamd-discuss |
|
From: Xiao, F. <Fu...@me...> - 2025-07-19 00:25:32
|
Hi Qianwei, I think the parameters of "timask1" and "scmask1" in your input file are a bit strange. These two parameters are the sequence numbers of the ligand in the amber format protein complex (pdb4amber command line can do this). Similar to the red box marked in the following picture (BEN is a small molecule). [cid:bc75a470-c323-4303-88c7-ed6455b0cb77] Small molecules are represented by an integer number, while peptides/antibodies are indicated by a sequence number such as xxx-xxx. In addition, iE = 2 can be deleted. Maybe you can try to check your parameters carefully. Best, Fu Xiao ________________________________ From: Miao, Yinglong <Yin...@me...> Sent: Friday, July 18, 2025 17:05 To: Li, Joyce <JO...@pi...> Cc: gamd-discuss <gam...@li...>; Xiao, Fu <Fu...@me...>; Joshi, Keya <ke...@em...> Subject: Re: [GaMD-discuss] Question about proper sigma0P value Hi Qianwei, I’m CC’ing Fu and Keya, experts in LiGaMD3 in our lab, and they may be able to help you more. 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 Editor-in-Chief npj Drug Discovery https://www.nature.com/npjdrugdiscov On Jul 17, 2025, at 6:36 PM, Li, Joyce via GaMD-discuss <gam...@li...> wrote: Hello, I am trying to use LiGAMD3 for measuring the binding kinetics between protein and ligand. The system I use is PDB 5JID. The original system has two ligands, but I kept only one for the GAMD. I followed the discussions in paper (10.1021/acs.jctc.4c00502) and tried to find the right sigma0p for unbinding the ligand. But even if I set the sigma0P to 8, the ligand still binds with the protein and the ligand RMSD is lower than 2 A after 100 ns production run. Attached are the codes for the simulation (pasted below as well) and the log file for the 100 ns production when P is 8. Did I do something wrong, or should I go for an even much higher P value? I have been trying this for months but not seenthe ligand to move. So I will appreciate any suggestions! Thank you, Best, Qianwei equbrilium cMD and GaMD equilbrium &cntrl ig = -1, ! Pseudo-random number generator (random seed based on time) ntx = 1, ! no initial coordinates and velocity(ASCII) imin = 0, ! Minimization (No) irest = 0, ! Restart the Simulation? (no) use previous velocity cut = 8, ! Non-bonding Cut-off (8 A) ntc = 2, ! SHAKE bond length constraints (constrain bonds with H) ntf = 1, ! compelete interaction is calculated tol = 0.000001, !relatve geometrical tolerance in shake dt = 0.002, ! timestep = 2fs iwrap = 1, !set for long runs to keep coordinate from overflowing nstlim = 44000000, ! Simulation length (sum of ntcmd and nteb total 88 ns) ntb = 2, ! PBC (Constant pressure) ntp = 1, ! isotropic position scaling barostat = 1, ! for control pressure, Berendsen ntt = 3, ! Temperature scaling (Stochastic version of Berendsen thermostat) gamma_ln=5.0, temp0 = 300, ! Reference temperature (300K, physiological temperature) tempi = 300.0, !Initial temperature ntpr = 500000, ! Energy save interval (every 500000 steps or 1 ns) ntwx = 500000, ! Coordinate/trajectory save interval (every 500000 steps or 1 ns) ntwprt = 7101, !number of atoms to be saved ntwr = 5000000, ! Write restart file every 10 ns igamd = 28, iE = 2, !thershold energy to the lower bound irest_gamd = 0, !not a restart gamd ntcmd = 4000000, !8 ns cMD nteb = 40000000, !80 ns GAMD ntave = 400000, !four times total number of atoms ntcmdprep = 800000, !1.6 ns number of cMD not collected for statistics calculation ntebprep = 800000, !1.6 ns GAMD for preperation sigma0P = 8.0, sigma0D = 6.0, sigma0B = 6.0, iEP=2, iED=2, iEB=1, bgpro2atm=1, edpro2atm=7076, gti_cpu_output=0, gti_add_sc=1, icfe = 1, ifsc = 1, timask1=':7077-7101', scmask1=':7077-7101', timask2='', scmask2='', &end END production GaMD production &cntrl ig = -1, ! Pseudo-random number generator (random seed based on time) ntx = 1, ! no initial coordinates and velocity(ASCII) imin = 0, ! Minimization (No) irest = 0, ! Restart the Simulation? (no) use previous velocity cut = 8, ! Non-bonding Cut-off (8 A) ntc = 2, ! SHAKE bond length constraints (constrain bonds with H) ntf = 1, ! compelete interaction is calculated tol = 0.000001, !relatve geometrical tolerance in shake dt = 0.002, ! timestep = 2fs iwrap = 1, !set for long runs to keep coordinate from overflowing nstlim = 50000000, ! Simulation length (sum of ntcmd and nteb total 100 ns) ntb = 2, ! PBC (Constant pressure) ntp = 1, ! isotropic position scaling barostat = 1, ! for control pressure, Berendsen ntt = 3, ! Temperature scaling (Stochastic version of Berendsen thermostat) gamma_ln=5.0, temp0 = 300, ! Reference temperature (300K, physiological temperature) tempi = 300.0, !Initial temperature ntpr = 250000, ! Energy save interval (every 250000 steps or 500ps) ntwx = 500, ! Coordinate/trajectory save interval (every 500 steps or 1ps) ntwprt = 7101, !number of atoms to be saved ntwr = 5000000, ! Write restart file every 10 ns igamd = 28, iE = 2, !thershold energy to the lower bound irest_gamd = 1, !a restart gamd ntcmd = 0, !no cMD nteb = 0, !0 GAMD ntave = 400000, !four times total number of atoms ntcmdprep = 0, !no cMD preperation ntebprep = 0, !noGAMD for preperation sigma0P = 8.0, sigma0D = 6.0, sigma0B = 6.0, iEP=2, iED=2, iEB=1, bgpro2atm=1, edpro2atm=7076, gti_cpu_output=0, gti_add_sc=1, icfe = 1, ifsc = 1, timask1=':7077-7101', scmask1=':7077-7101', timask2='', scmask2='', &end END _______________________________________________ GaMD-discuss mailing list GaM...@li...<mailto:GaM...@li...> https://lists.sourceforge.net/lists/listinfo/gamd-discuss |
|
From: Miao, Y. <Yin...@me...> - 2025-07-18 21:20:16
|
Hi Qianwei, I’m CC’ing Fu and Keya, experts in LiGaMD3 in our lab, and they may be able to help you more. 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 Editor-in-Chief npj Drug Discovery https://www.nature.com/npjdrugdiscov On Jul 17, 2025, at 6:36 PM, Li, Joyce via GaMD-discuss <gam...@li...> wrote: Hello, I am trying to use LiGAMD3 for measuring the binding kinetics between protein and ligand. The system I use is PDB 5JID. The original system has two ligands, but I kept only one for the GAMD. I followed the discussions in paper (10.1021/acs.jctc.4c00502) and tried to find the right sigma0p for unbinding the ligand. But even if I set the sigma0P to 8, the ligand still binds with the protein and the ligand RMSD is lower than 2 A after 100 ns production run. Attached are the codes for the simulation (pasted below as well) and the log file for the 100 ns production when P is 8. Did I do something wrong, or should I go for an even much higher P value? I have been trying this for months but not seenthe ligand to move. So I will appreciate any suggestions! Thank you, Best, Qianwei equbrilium cMD and GaMD equilbrium &cntrl ig = -1, ! Pseudo-random number generator (random seed based on time) ntx = 1, ! no initial coordinates and velocity(ASCII) imin = 0, ! Minimization (No) irest = 0, ! Restart the Simulation? (no) use previous velocity cut = 8, ! Non-bonding Cut-off (8 A) ntc = 2, ! SHAKE bond length constraints (constrain bonds with H) ntf = 1, ! compelete interaction is calculated tol = 0.000001, !relatve geometrical tolerance in shake dt = 0.002, ! timestep = 2fs iwrap = 1, !set for long runs to keep coordinate from overflowing nstlim = 44000000, ! Simulation length (sum of ntcmd and nteb total 88 ns) ntb = 2, ! PBC (Constant pressure) ntp = 1, ! isotropic position scaling barostat = 1, ! for control pressure, Berendsen ntt = 3, ! Temperature scaling (Stochastic version of Berendsen thermostat) gamma_ln=5.0, temp0 = 300, ! Reference temperature (300K, physiological temperature) tempi = 300.0, !Initial temperature ntpr = 500000, ! Energy save interval (every 500000 steps or 1 ns) ntwx = 500000, ! Coordinate/trajectory save interval (every 500000 steps or 1 ns) ntwprt = 7101, !number of atoms to be saved ntwr = 5000000, ! Write restart file every 10 ns igamd = 28, iE = 2, !thershold energy to the lower bound irest_gamd = 0, !not a restart gamd ntcmd = 4000000, !8 ns cMD nteb = 40000000, !80 ns GAMD ntave = 400000, !four times total number of atoms ntcmdprep = 800000, !1.6 ns number of cMD not collected for statistics calculation ntebprep = 800000, !1.6 ns GAMD for preperation sigma0P = 8.0, sigma0D = 6.0, sigma0B = 6.0, iEP=2, iED=2, iEB=1, bgpro2atm=1, edpro2atm=7076, gti_cpu_output=0, gti_add_sc=1, icfe = 1, ifsc = 1, timask1=':7077-7101', scmask1=':7077-7101', timask2='', scmask2='', &end END production GaMD production &cntrl ig = -1, ! Pseudo-random number generator (random seed based on time) ntx = 1, ! no initial coordinates and velocity(ASCII) imin = 0, ! Minimization (No) irest = 0, ! Restart the Simulation? (no) use previous velocity cut = 8, ! Non-bonding Cut-off (8 A) ntc = 2, ! SHAKE bond length constraints (constrain bonds with H) ntf = 1, ! compelete interaction is calculated tol = 0.000001, !relatve geometrical tolerance in shake dt = 0.002, ! timestep = 2fs iwrap = 1, !set for long runs to keep coordinate from overflowing nstlim = 50000000, ! Simulation length (sum of ntcmd and nteb total 100 ns) ntb = 2, ! PBC (Constant pressure) ntp = 1, ! isotropic position scaling barostat = 1, ! for control pressure, Berendsen ntt = 3, ! Temperature scaling (Stochastic version of Berendsen thermostat) gamma_ln=5.0, temp0 = 300, ! Reference temperature (300K, physiological temperature) tempi = 300.0, !Initial temperature ntpr = 250000, ! Energy save interval (every 250000 steps or 500ps) ntwx = 500, ! Coordinate/trajectory save interval (every 500 steps or 1ps) ntwprt = 7101, !number of atoms to be saved ntwr = 5000000, ! Write restart file every 10 ns igamd = 28, iE = 2, !thershold energy to the lower bound irest_gamd = 1, !a restart gamd ntcmd = 0, !no cMD nteb = 0, !0 GAMD ntave = 400000, !four times total number of atoms ntcmdprep = 0, !no cMD preperation ntebprep = 0, !noGAMD for preperation sigma0P = 8.0, sigma0D = 6.0, sigma0B = 6.0, iEP=2, iED=2, iEB=1, bgpro2atm=1, edpro2atm=7076, gti_cpu_output=0, gti_add_sc=1, icfe = 1, ifsc = 1, timask1=':7077-7101', scmask1=':7077-7101', timask2='', scmask2='', &end END _______________________________________________ GaMD-discuss mailing list GaM...@li...<mailto:GaM...@li...> https://lists.sourceforge.net/lists/listinfo/gamd-discuss |
|
From: Li, J. <JO...@pi...> - 2025-07-18 03:11:02
|
Hello, I am trying to use LiGAMD3 for measuring the binding kinetics between protein and ligand. The system I use is PDB 5JID. The original system has two ligands, but I kept only one for the GAMD. I followed the discussions in paper (10.1021/acs.jctc.4c00502) and tried to find the right sigma0p for unbinding the ligand. But even if I set the sigma0P to 8, the ligand still binds with the protein and the ligand RMSD is lower than 2 A after 100 ns production run. Attached are the codes for the simulation (pasted below as well) and the log file for the 100 ns production when P is 8. Did I do something wrong, or should I go for an even much higher P value? I have been trying this for months but not seenthe ligand to move. So I will appreciate any suggestions! Thank you, Best, Qianwei equbrilium cMD and GaMD equilbrium &cntrl ig = -1, ! Pseudo-random number generator (random seed based on time) ntx = 1, ! no initial coordinates and velocity(ASCII) imin = 0, ! Minimization (No) irest = 0, ! Restart the Simulation? (no) use previous velocity cut = 8, ! Non-bonding Cut-off (8 A) ntc = 2, ! SHAKE bond length constraints (constrain bonds with H) ntf = 1, ! compelete interaction is calculated tol = 0.000001, !relatve geometrical tolerance in shake dt = 0.002, ! timestep = 2fs iwrap = 1, !set for long runs to keep coordinate from overflowing nstlim = 44000000, ! Simulation length (sum of ntcmd and nteb total 88 ns) ntb = 2, ! PBC (Constant pressure) ntp = 1, ! isotropic position scaling barostat = 1, ! for control pressure, Berendsen ntt = 3, ! Temperature scaling (Stochastic version of Berendsen thermostat) gamma_ln=5.0, temp0 = 300, ! Reference temperature (300K, physiological temperature) tempi = 300.0, !Initial temperature ntpr = 500000, ! Energy save interval (every 500000 steps or 1 ns) ntwx = 500000, ! Coordinate/trajectory save interval (every 500000 steps or 1 ns) ntwprt = 7101, !number of atoms to be saved ntwr = 5000000, ! Write restart file every 10 ns igamd = 28, iE = 2, !thershold energy to the lower bound irest_gamd = 0, !not a restart gamd ntcmd = 4000000, !8 ns cMD nteb = 40000000, !80 ns GAMD ntave = 400000, !four times total number of atoms ntcmdprep = 800000, !1.6 ns number of cMD not collected for statistics calculation ntebprep = 800000, !1.6 ns GAMD for preperation sigma0P = 8.0, sigma0D = 6.0, sigma0B = 6.0, iEP=2, iED=2, iEB=1, bgpro2atm=1, edpro2atm=7076, gti_cpu_output=0, gti_add_sc=1, icfe = 1, ifsc = 1, timask1=':7077-7101', scmask1=':7077-7101', timask2='', scmask2='', &end END production GaMD production &cntrl ig = -1, ! Pseudo-random number generator (random seed based on time) ntx = 1, ! no initial coordinates and velocity(ASCII) imin = 0, ! Minimization (No) irest = 0, ! Restart the Simulation? (no) use previous velocity cut = 8, ! Non-bonding Cut-off (8 A) ntc = 2, ! SHAKE bond length constraints (constrain bonds with H) ntf = 1, ! compelete interaction is calculated tol = 0.000001, !relatve geometrical tolerance in shake dt = 0.002, ! timestep = 2fs iwrap = 1, !set for long runs to keep coordinate from overflowing nstlim = 50000000, ! Simulation length (sum of ntcmd and nteb total 100 ns) ntb = 2, ! PBC (Constant pressure) ntp = 1, ! isotropic position scaling barostat = 1, ! for control pressure, Berendsen ntt = 3, ! Temperature scaling (Stochastic version of Berendsen thermostat) gamma_ln=5.0, temp0 = 300, ! Reference temperature (300K, physiological temperature) tempi = 300.0, !Initial temperature ntpr = 250000, ! Energy save interval (every 250000 steps or 500ps) ntwx = 500, ! Coordinate/trajectory save interval (every 500 steps or 1ps) ntwprt = 7101, !number of atoms to be saved ntwr = 5000000, ! Write restart file every 10 ns igamd = 28, iE = 2, !thershold energy to the lower bound irest_gamd = 1, !a restart gamd ntcmd = 0, !no cMD nteb = 0, !0 GAMD ntave = 400000, !four times total number of atoms ntcmdprep = 0, !no cMD preperation ntebprep = 0, !noGAMD for preperation sigma0P = 8.0, sigma0D = 6.0, sigma0B = 6.0, iEP=2, iED=2, iEB=1, bgpro2atm=1, edpro2atm=7076, gti_cpu_output=0, gti_add_sc=1, icfe = 1, ifsc = 1, timask1=':7077-7101', scmask1=':7077-7101', timask2='', scmask2='', &end END |
|
From: Miao, Y. <Yin...@me...> - 2025-07-13 20:15:04
|
Hi Arpan,
Attached I’m sending you a copy of the Pep-GaMD simulation input files we used, with the .in files pasted here:
::::::::::::::
md.in
::::::::::::::
GaMD equilibration simulation
&cntrl
imin=0, ! No minimization
irest=0, ! This IS a new MD simulation
ntx=1, ! read coordinates only
! Temperature control
ntt=3, ! Langevin dynamics
gamma_ln=1.0, ! Friction coefficient (ps^-1)
tempi=300.0, ! Initial temperature
temp0=300.0, ! Target temperature
ig=-1, ! random seed
! Potential energy control
cut=9.0, ! nonbonded cutoff, in Angstroms
! MD settings
nstlim=32800000, ! simulation length
dt=0.002, ! time step (ps)
! SHAKE
ntc=2, ! Constrain bonds containing hydrogen
ntf=1, ! Do not calculate forces of bonds containing hydrogen
! Control how often information is printed
ntpr=100, ! Print energies every 1000 steps
ntwx=100, ! Print coordinates every 1000 steps to the trajectory
ntwr=5000, ! Print a restart file every 10K steps (can be less frequent)
! ntwv=-1, ! Uncomment to also print velocities to trajectory
! ntwf=-1, ! Uncomment to also print forces to trajectory
ntxo=2, ! Write NetCDF format
ioutfm=1, ! Write NetCDF format (always do this!)
! Wrap coordinates when printing them to the same unit cell
iwrap=1,
ntwprt = 1087,
! Constant pressure control. Note that ntp=3 requires barostat=1
barostat=1, ! Berendsen... change to 2 for MC barostat
ntp=1, ! 1=isotropic, 2=anisotropic, 3=semi-isotropic w/ surften
pres0=1.0, ! Target external pressure, in bar
taup=0.5, ! Berendsen coupling constant (ps)
! Constant surface tension (needed for semi-isotropic scaling). Uncomment
! for this feature. csurften must be nonzero if ntp=3 above
! csurften=3, ! Interfaces in 1=yz plane, 2=xz plane, 3=xy plane
! gamma_ten=0.0, ! Surface tension (dyne/cm). 0 gives pure semi-iso scaling
! ninterface=2, ! Number of interfaces (2 for bilayer)
! Set water atom/residue names for SETTLE recognition
!watnam='TIP3', ! Water residues are named TIP3
!owtnm='OH2', ! Water oxygens are named OH2
! GaMD parameters
igamd = 15, iEP = 2,iED=1,irest_gamd = 0,
ntcmd = 800000, nteb = 32000000, ntave = 160000,
ntcmdprep = 320000, ntebprep = 320000,
sigma0P = 4.0, sigma0D = 6.0,
gti_cpu_output = 0, gti_add_sc = 1,
icfe = 1,
ifsc = 1,
timask1 = ':58-68',
scmask1 = ':58-68',
timask2 = '',
scmask2='',
/
::::::::::::::
gamd-restart.in
::::::::::::::
GaMD production simulation
&cntrl
imin=0, ! No minimization
irest=0, ! This IS a new MD simulation
ntx=1, ! read coordinates only
! Temperature control
ntt=3, ! Langevin dynamics
gamma_ln=1.0, ! Friction coefficient (ps^-1)
tempi=300.0, ! Initial temperature
temp0=300.0, ! Target temperature
ig=-1, ! random seed
! Potential energy control
cut=9.0, ! nonbonded cutoff, in Angstroms
! MD settings
nstlim=50000000, ! simulation length
dt=0.002, ! time step (ps)
! SHAKE
ntc=2, ! Constrain bonds containing hydrogen
ntf=1, ! Do not calculate forces of bonds containing hydrogen
! Control how often information is printed
ntpr=100, ! Print energies every 1000 steps
ntwx=100, ! Print coordinates every 1000 steps to the trajectory
ntwr=5000, ! Print a restart file every 10K steps (can be less frequent)
! ntwv=-1, ! Uncomment to also print velocities to trajectory
! ntwf=-1, ! Uncomment to also print forces to trajectory
ntxo=2, ! Write NetCDF format
ioutfm=1, ! Write NetCDF format (always do this!)
! Wrap coordinates when printing them to the same unit cell
iwrap=1,
ntwprt = 1087,
! Constant pressure control. Note that ntp=3 requires barostat=1
barostat=1, ! Berendsen... change to 2 for MC barostat
ntp=1, ! 1=isotropic, 2=anisotropic, 3=semi-isotropic w/ surften
pres0=1.0, ! Target external pressure, in bar
taup=0.5, ! Berendsen coupling constant (ps)
! Constant surface tension (needed for semi-isotropic scaling). Uncomment
! for this feature. csurften must be nonzero if ntp=3 above
! csurften=3, ! Interfaces in 1=yz plane, 2=xz plane, 3=xy plane
! gamma_ten=0.0, ! Surface tension (dyne/cm). 0 gives pure semi-iso scaling
! ninterface=2, ! Number of interfaces (2 for bilayer)
! Set water atom/residue names for SETTLE recognition
!watnam='TIP3', ! Water residues are named TIP3
!owtnm='OH2', ! Water oxygens are named OH2
! GaMD parameters
igamd = 15, iEP = 2,iED=1, irest_gamd = 1,
ntcmd = 0, nteb = 0, ntave = 160000,
ntcmdprep = 0, ntebprep = 0,
sigma0P = 4.0, sigma0D = 6.0,
gti_cpu_output = 0, gti_add_sc = 1,
icfe = 1,
ifsc = 1,
timask1 = ':58-68',
scmask1 = ':58-68',
timask2 = '',
scmask2='',
/
::::::::::::::
md-restart.in
::::::::::::::
GaMD restart simulation
&cntrl
imin=0, ! No minimization
irest=1, ! This IS a new MD simulation
ntx=5, ! read coordinates only
! Temperature control
ntt=3, ! Langevin dynamics
gamma_ln=1.0, ! Friction coefficient (ps^-1)
tempi=300.0, ! Initial temperature
temp0=300.0, ! Target temperature
ig=-1, ! random seed
! Potential energy control
cut=9.0, ! nonbonded cutoff, in Angstroms
! MD settings
nstlim=50000000, ! simulation length
dt=0.002, ! time step (ps)
! SHAKE
ntc=2, ! Constrain bonds containing hydrogen
ntf=1, ! Do not calculate forces of bonds containing hydrogen
! Control how often information is printed
ntpr=100, ! Print energies every 1000 steps
ntwx=100, ! Print coordinates every 1000 steps to the trajectory
ntwr=5000, ! Print a restart file every 10K steps (can be less frequent)
! ntwv=-1, ! Uncomment to also print velocities to trajectory
! ntwf=-1, ! Uncomment to also print forces to trajectory
ntxo=2, ! Write NetCDF format
ioutfm=1, ! Write NetCDF format (always do this!)
! Wrap coordinates when printing them to the same unit cell
iwrap=1,
ntwprt = 1087,
! Constant pressure control. Note that ntp=3 requires barostat=1
barostat=1, ! Berendsen... change to 2 for MC barostat
ntp=1, ! 1=isotropic, 2=anisotropic, 3=semi-isotropic w/ surften
pres0=1.0, ! Target external pressure, in bar
taup=0.5, ! Berendsen coupling constant (ps)
! Constant surface tension (needed for semi-isotropic scaling). Uncomment
! for this feature. csurften must be nonzero if ntp=3 above
! csurften=3, ! Interfaces in 1=yz plane, 2=xz plane, 3=xy plane
! gamma_ten=0.0, ! Surface tension (dyne/cm). 0 gives pure semi-iso scaling
! ninterface=2, ! Number of interfaces (2 for bilayer)
! Set water atom/residue names for SETTLE recognition
!watnam='TIP3', ! Water residues are named TIP3
!owtnm='OH2', ! Water oxygens are named OH2
! GaMD parameters
igamd = 15, iEP = 2,iED=1, irest_gamd = 1,
ntcmd = 0, nteb = 0, ntave = 160000,
ntcmdprep = 0, ntebprep = 0,
sigma0P = 4.0, sigma0D = 6.0,
gti_cpu_output = 0, gti_add_sc = 1,
icfe = 1,
ifsc = 1,
timask1 = ':58-68',
scmask1 = ':58-68',
timask2 = '',
scmask2='',
/
The simulations may run as follows:
##GaMD equilibration
#~/bin/pmemd.cuda -O -i md.in -p sys.prm -c density.rst -o md.out -r gamd-1.rst -x md-1.nc -gamd gamd-1.log
##the GaMD production
~/bin/pmemd.cuda -O -i gamd-restart.in -p complex.prmtop -c gamd-1.rst -o gamd-2.out -x md-2.nc -r gamd-2.rst -gamd gamd-2.log
##GaMD restart production
~/bin/pmemd.cuda -O -i md-restart.in -p complex.prmtop -c gamd-2.rst -o gamd-3.out -x md-3.nc -r gamd-3.rst -gamd gamd-3.log
for i in `seq 4 15`
do
j=`echo $i|awk '{print $1-1}'`
~/bin/pmemd.cuda -O -i md-restart.in -p complex.prmtop -c gamd-$j.rst -o gamd-$i.out -x md-$i.nc -r gamd-$i.rst -gamd gamd-$i.log
done
Hope they help!
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 Jul 11, 2025, at 2:11 PM, Arpan Tapdiya <agt...@nc...> wrote:
You don't often get email from agt...@nc.... Learn why this is important<https://aka.ms/LearnAboutSenderIdentification>
Hello,
I am Arpan Tapdiya, a PhD student in the Prof.Carol Hall lab at NC State University, Raleigh. I am trying to replicate Pep-GaMD method for enhanced sampling (reference article - https://doi.org/10.1063/5.0021399). While trying to run the three production runs independently for the system with PDB ID - 1CKA, I ran into a problem where the GaMD parameters are not being updated in the game-restart.dat file.
I have the production run for 1microsecond broken down into 50 parts (each of 20 nanoseconds). My game-restart.dat file has not been updated since the equilibration part. My parameters for the production and equilibration runs are attached, and the loop for the production run is being executed through the following commands -
##syntax starts here -
prev_rst="equil.rst"
for i in {1..50}; do
out="prod${i}.out"
rst="prod${i}.rst"
traj="prod${i}.nc"
gamd_data="gamd_prod${i}.dat"
echo "Running production step $i..."
${pmemd_path}/pmemd.cuda_SPFP -O -i prod.in<http://prod.in/> -o $out -p pep.prmtop -c $prev_rst -r $rst -x $traj -gamd gamd-restart.dat "${gamd_data}"
prev_rst=$rst
cp gamd-restart.dat gamd-restart{i}.dat
echo "Production step ${i} completed"
done
Can you help me solve the issue where my GaMD parameters are updated after every production step such that the simulation has a continuous bias potential?
Thank you.
Best,
Arpan
--
Arpan Tapdiya
Graduate Student
Department of Chemical and Biomolecular Engineering
North Carolina State University
<equil.in><prod.in>
|
|
From: Arpan T. <agt...@nc...> - 2025-07-11 19:21:50
|
Hello, I am Arpan Tapdiya, a PhD student in the Prof.Carol Hall lab at NC State University, Raleigh. I am trying to replicate Pep-GaMD method for enhanced sampling (reference article - https://doi.org/10.1063/5.0021399). While trying to run the three production runs independently for the system with PDB ID - 1CKA, I ran into a problem where the GaMD parameters are not being updated in the game-restart.dat file. I have the production run for 1microsecond broken down into 50 parts (each of 20 nanoseconds). My game-restart.dat file has not been updated since the equilibration part. My parameters for the production and equilibration runs are attached, and the loop for the production run is being executed through the following commands - ##syntax starts here - prev_rst="equil.rst" for i in {1..50}; do out="prod${i}.out" rst="prod${i}.rst" traj="prod${i}.nc" gamd_data="gamd_prod${i}.dat" echo "Running production step $i..." ${pmemd_path}/pmemd.cuda_SPFP -O -i prod.in -o $out -p pep.prmtop -c $prev_rst -r $rst -x $traj -gamd gamd-restart.dat "${gamd_data}" prev_rst=$rst cp gamd-restart.dat gamd-restart{i}.dat echo "Production step ${i} completed" done Can you help me solve the issue where my GaMD parameters are updated after every production step such that the simulation has a continuous bias potential? Thank you. Best, Arpan -- Arpan Tapdiya Graduate Student Department of Chemical and Biomolecular Engineering North Carolina State University |
|
From: Miao, Y. <Yin...@me...> - 2025-06-24 13:48:59
|
Hi Irene, Here is what we have: “Increase MaxNumberTIAtom to a value that exceeds the maximum number of TI atoms specified in TIMASK1 and SCMASK1, according to your system’s requirements.” TIMASK1 and SCMAKS1 correspond to the binding “ligand” I think. 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 Jun 24, 2025, at 8:56 AM, Irene Maffucci <ire...@ut...> wrote: Dear GaMD users, Is the MaxNumberTIAtom need to be fixed as greater than the sum of the number of the atoms of the two binding partners? Is there an upper limit? Best regards Irene --------------------------------------------------------------------- Irene Maffucci, PhD Lab. de Génie Enzymatique et Cellulaire UMR 7025 Université de technologie de Compiègne Centre de Recherche Royallieu CS60319 60203 Compiègne cedex Tel. +33 (0)3 44 23 43 42 Le 30 mai 2025 à 16:12, Miao, Yinglong <Yin...@me...> a écrit : Hi Irene, That shouldn’t be an issue. 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 May 30, 2025, at 4:04 AM, Irene Maffucci <ire...@ut...> wrote: Hi Yinglong, Thanks for the details. Is there an upper limit for the maximum number of TI atoms ? Would 4000 atoms be an issue ? Thanks for the help Best Irene --------------------------------------------------------------------- Irene Maffucci, PhD Lab. de Génie Enzymatique et Cellulaire UMR 7025 Université de technologie de Compiègne Centre de Recherche Royallieu CS60319 60203 Compiègne cedex Tel. +33 (0)3 44 23 43 42 Le 23 mai 2025 à 20:24, Miao, Yinglong <Yin...@me...> a écrit : Hi Irene, I’m glad you debugged an issue! I have two additional questions: first of all, in order to set the best sigma value, one should just check that the k_0 reaches 1, without having high sigma_DeltaV, therefore, if one uses the lower bound condition, when k_0 reaches 1, increasing the sigma value, while keeping sigma_DeltaV < 10k_BT, does not have any effect on the potential boost, am I right? Therefore, one should keep the sigma to the minimum value allowing to reach k_0 = 1 to have a proper reweighing, right? Right. And normally the default values of sigma0 should be able to make k0 reach 1 before the end of GaMD equilibration. My second question actually concerns the timask1, scmask1 parameters with igamd=16 or 17: if the protein is big and the atom limit is overcome, would it be possible to specify a subset of atoms (i.e. backbone atoms)? No, since you want to boot the non-bonded interaction energy, all related atoms need to be selected. Jinan in our lab came up a fix for the atom limit issue: https://www.med.unc.edu/pharm/miaolab/conferences/gamd2024/faq/ Q: How to fix limitation of the maximum number of TI atoms for running PPI-GaMD simulations? A: To address the limitation on the maximum number of TI atoms, adjust the MaxNumberTIAtom parameter in the ${AMBERHOME}/src/pmemd/src/gti_controlVariable.i file. Increase MaxNumberTIAtom to a value that exceeds the maximum number of TI atoms specified in TIMASK1 and SCMASK1, according to your system’s requirements. After making these adjustments, recompile the AMBER24 code with the updated gti_controlVariable.i file. Hope this helps, 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 May 22, 2025, at 3:18 AM, Irene Maffucci <ire...@ut...> wrote: Hello Yinglong, I checked my input file and I have actually made a mistake in the timask1, thank you for letting me notice it! I have two additional questions: first of all, in order to set the best sigma value, one should just check that the k_0 reaches 1, without having high sigma_DeltaV, therefore, if one uses the lower bound condition, when k_0 reaches 1, increasing the sigma value, while keeping sigma_DeltaV < 10k_BT, does not have any effect on the potential boost, am I right? Therefore, one should keep the sigma to the minimum value allowing to reach k_0 = 1 to have a proper reweighing, right? My second question actually concerns the timask1, scmask1 parameters with igamd=16 or 17: if the protein is big and the atom limit is overcome, would it be possible to specify a subset of atoms (i.e. backbone atoms)? Thanks in advance for the help Best Irene --------------------------------------------------------------------- Irene Maffucci, PhD Lab. de Génie Enzymatique et Cellulaire UMR 7025 Université de technologie de Compiègne Centre de Recherche Royallieu CS60319 60203 Compiègne cedex Tel. +33 (0)3 44 23 43 42 Le 22 mai 2025 à 03:07, Miao, Yinglong <Yin...@me...> a écrit : Hi Irene, Great that you observed unbinding/binding events of the oligonucleotide in your simulations! As you read about the PPI-GaMD algorithm, it applies selective boost potential to the non-bonded interaction potential energy of two binding partners, which can be proteins and probably other biomolecules of your interest like protein-oligonucleotide. I guess it’s only a naming issue here. With igamd=17, the dual-boost mode is indeed turned on. You may want to double check if parameters such as timask1, scmask1, bgpro2atm and edpro2atm are set properly so that the boosted potential energies are calculated correctly. If you can perhaps share a copy of the mdout file for the “Equilibration” part, we can look into the issue more. 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 May 20, 2025, at 10:00 AM, Irene Maffucci <ire...@ut...> wrote: Hello everyone, I am running PPI-GaMD on a protein-oligonucleotide complex. First of all, although PPI-GaMD has been developed for PPIs, can it be applied also on the kind of complex I am simulating? I am observing unbinding/binding events of the oligonucleotide, but I would like to know your opinion about the suitability of GaMD for these systems. I am running my simulations with Amber by setting igamd=17, with therefore a dual boost. However, when I give a look at the log files I have the statistics for VmaxD, VminD, VavgD, sigmaVD, but all the statistics for the VmaxP, VminP, VavgP and sigmaVP are 0. Is this normal? In addition, for the choice of the sigma0 values, I let them to 6 kcal/mol and I verified that the k0 reached 1, but I observe a large sigmaVD. What should I do in this case? Thanks and best regards Irene --------------------------------------------------------------------- Irene Maffucci, PhD Lab. de Génie Enzymatique et Cellulaire UMR 7025 Université de technologie de Compiègne Centre de Recherche Royallieu CS60319 60203 Compiègne cedex Tel. +33 (0)3 44 23 43 42 _______________________________________________ GaMD-discuss mailing list GaM...@li... https://lists.sourceforge.net/lists/listinfo/gamd-discuss |
|
From: Irene M. <ire...@ut...> - 2025-06-24 12:56:29
|
Dear GaMD users, Is the MaxNumberTIAtom need to be fixed as greater than the sum of the number of the atoms of the two binding partners? Is there an upper limit? Best regards Irene --------------------------------------------------------------------- Irene Maffucci, PhD Lab. de Génie Enzymatique et Cellulaire UMR 7025 Université de technologie de Compiègne Centre de Recherche Royallieu CS60319 60203 Compiègne cedex Tel. +33 (0)3 44 23 43 42 > Le 30 mai 2025 à 16:12, Miao, Yinglong <Yin...@me...> a écrit : > > Hi Irene, > > That shouldn’t be an issue. > > 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 May 30, 2025, at 4:04 AM, Irene Maffucci <ire...@ut...> wrote: >> >> Hi Yinglong, >> >> Thanks for the details. >> Is there an upper limit for the maximum number of TI atoms ? Would 4000 atoms be an issue ? >> >> Thanks for the help >> Best >> Irene >> --------------------------------------------------------------------- >> Irene Maffucci, PhD >> Lab. de Génie Enzymatique et Cellulaire UMR 7025 >> Université de technologie de Compiègne >> Centre de Recherche Royallieu >> CS60319 >> 60203 Compiègne cedex >> Tel. +33 (0)3 44 23 43 42 >> >>> Le 23 mai 2025 à 20:24, Miao, Yinglong <Yin...@me...> a écrit : >>> >>> Hi Irene, >>> >>> I’m glad you debugged an issue! >>> >>>> I have two additional questions: first of all, in order to set the best sigma value, one should just check that the k_0 reaches 1, without having high sigma_DeltaV, therefore, if one uses the lower bound condition, when k_0 reaches 1, increasing the sigma value, while keeping sigma_DeltaV < 10k_BT, does not have any effect on the potential boost, am I right? Therefore, one should keep the sigma to the minimum value allowing to reach k_0 = 1 to have a proper reweighing, right? >>> Right. And normally the default values of sigma0 should be able to make k0 reach 1 before the end of GaMD equilibration. >>> >>>> My second question actually concerns the timask1, scmask1 parameters with igamd=16 or 17: if the protein is big and the atom limit is overcome, would it be possible to specify a subset of atoms (i.e. backbone atoms)? >>> No, since you want to boot the non-bonded interaction energy, all related atoms need to be selected. Jinan in our lab came up a fix for the atom limit issue: https://www.med.unc.edu/pharm/miaolab/conferences/gamd2024/faq/ >>> >>> Q: How to fix limitation of the maximum number of TI atoms for running PPI-GaMD simulations? >>> A: To address the limitation on the maximum number of TI atoms, adjust the MaxNumberTIAtom parameter in the ${AMBERHOME}/src/pmemd/src/gti_controlVariable.i file. Increase MaxNumberTIAtom to a value that exceeds the maximum number of TI atoms specified in TIMASK1 and SCMASK1, according to your system’s requirements. After making these adjustments, recompile the AMBER24 code with the updated gti_controlVariable.i file. >>> Hope this helps, >>> 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 May 22, 2025, at 3:18 AM, Irene Maffucci <ire...@ut...> wrote: >>>> >>>> Hello Yinglong, >>>> >>>> I checked my input file and I have actually made a mistake in the timask1, thank you for letting me notice it! >>>> I have two additional questions: first of all, in order to set the best sigma value, one should just check that the k_0 reaches 1, without having high sigma_DeltaV, therefore, if one uses the lower bound condition, when k_0 reaches 1, increasing the sigma value, while keeping sigma_DeltaV < 10k_BT, does not have any effect on the potential boost, am I right? Therefore, one should keep the sigma to the minimum value allowing to reach k_0 = 1 to have a proper reweighing, right? >>>> My second question actually concerns the timask1, scmask1 parameters with igamd=16 or 17: if the protein is big and the atom limit is overcome, would it be possible to specify a subset of atoms (i.e. backbone atoms)? >>>> >>>> Thanks in advance for the help >>>> Best >>>> Irene >>>> --------------------------------------------------------------------- >>>> Irene Maffucci, PhD >>>> Lab. de Génie Enzymatique et Cellulaire UMR 7025 >>>> Université de technologie de Compiègne >>>> Centre de Recherche Royallieu >>>> CS60319 >>>> 60203 Compiègne cedex >>>> Tel. +33 (0)3 44 23 43 42 >>>> >>>>> Le 22 mai 2025 à 03:07, Miao, Yinglong <Yin...@me...> a écrit : >>>>> >>>>> Hi Irene, >>>>> >>>>> Great that you observed unbinding/binding events of the oligonucleotide in your simulations! As you read about the PPI-GaMD algorithm, it applies selective boost potential to the non-bonded interaction potential energy of two binding partners, which can be proteins and probably other biomolecules of your interest like protein-oligonucleotide. I guess it’s only a naming issue here. >>>>> >>>>> With igamd=17, the dual-boost mode is indeed turned on. You may want to double check if parameters such as timask1, scmask1, bgpro2atm and edpro2atm are set properly so that the boosted potential energies are calculated correctly. >>>>> >>>>> If you can perhaps share a copy of the mdout file for the “Equilibration” part, we can look into the issue more. >>>>> >>>>> 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 May 20, 2025, at 10:00 AM, Irene Maffucci <ire...@ut...> wrote: >>>>>> >>>>>> Hello everyone, >>>>>> >>>>>> I am running PPI-GaMD on a protein-oligonucleotide complex. >>>>>> First of all, although PPI-GaMD has been developed for PPIs, can it be applied also on the kind of complex I am simulating? I am observing unbinding/binding events of the oligonucleotide, but I would like to know your opinion about the suitability of GaMD for these systems. >>>>>> I am running my simulations with Amber by setting igamd=17, with therefore a dual boost. However, when I give a look at the log files I have the statistics for VmaxD, VminD, VavgD, sigmaVD, but all the statistics for the VmaxP, VminP, VavgP and sigmaVP are 0. Is this normal? >>>>>> In addition, for the choice of the sigma0 values, I let them to 6 kcal/mol and I verified that the k0 reached 1, but I observe a large sigmaVD. What should I do in this case? >>>>>> >>>>>> Thanks and best regards >>>>>> Irene >>>>>> --------------------------------------------------------------------- >>>>>> Irene Maffucci, PhD >>>>>> Lab. de Génie Enzymatique et Cellulaire UMR 7025 >>>>>> Université de technologie de Compiègne >>>>>> Centre de Recherche Royallieu >>>>>> CS60319 >>>>>> 60203 Compiègne cedex >>>>>> Tel. +33 (0)3 44 23 43 42 >>>>>> >>>>>> _______________________________________________ >>>>>> GaMD-discuss mailing list >>>>>> GaM...@li... >>>>>> https://lists.sourceforge.net/lists/listinfo/gamd-discuss >>>>> >>>> >>> >> > |
|
From: Miao, Y. <Yin...@me...> - 2025-06-24 03:05:10
|
Hi Christine, You may test setting a larger number for “accelMDGStatWindow” first, and then depending on what you observe in the equilibration outputs, you can probably shorten the equilibration simulation. Note that it’s more meaningful to analyze the boost potential distribution from the GaMD production simulations. 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 Editor-in-Chief npj Drug Discovery https://www.nature.com/npjdrugdiscov ======= For discussion on GaMD, please subscribe to mailing list: https://sourceforge.net/projects/gamd/lists/gamd-discuss, post emails to "gam...@li..." and help others as much as you can! ======= On Jun 23, 2025, at 2:43 PM, Christine Qian <chr...@uc...> wrote: You don't often get email from chr...@uc...<mailto:chr...@uc...>. Learn why this is important<https://aka.ms/LearnAboutSenderIdentification> Hi Yinglong, Thank you for your help! Just to confirm, do I either shorten my equilibrium simulation or set a larger number for "accelMDGStatWindow," or can I do both to my parameters? Also, in my earlier email, I forgot to attach some of the graphs that I plotted to check whether my systems are equilibrated. I've attached them here now. Would you say that the system is behaving as expected and is equilibrated? Thank you very much once again for your time and help! All the best, Christine ________________________________ From: Miao, Yinglong <Yin...@me...> Sent: Friday, June 20, 2025 1:50 PM To: Christine Qian <chr...@uc...> Cc: gamd-discuss <gam...@li...> Subject: Re: [GaMD-discuss] GaMD Equilibration Time Query Hi Christine, It seems ok to shorten your equilibration simulation. Meanwhile, you may set a larger number for “accelMDGStatWindow”, which could be roughly 4 times the total number of atoms in your system. 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 Jun 20, 2025, at 1:52 PM, Christine Qian <chr...@uc...> wrote: You don't often get email from chr...@uc...<mailto:chr...@uc...>. Learn why this is important<https://aka.ms/LearnAboutSenderIdentification> Dear GaMD developers, I am running GaMD on several protein (a monomer of 42 residues)-membrane systems with waters and ions. The systems have sizes of about 170000 atoms. This is the parameters that I'm using for my equilibration: # GaMD Equilibration Params _________________________ accelMD on accelMDdual on accelMDdihe on accelMDG on accelMDGiE 1 accelMDGRestart off accelMDGcMDPrepSteps 200000 accelMDGEquiPrepSteps 200000 accelMDGStatWindow 20000 accelMDGcMDSteps 1000000 accelMDGEquiSteps 25000000 accelMDOutFreq 1000 accelMDGsigma0P 6.0 accelMDGsigma0D 6.0 run 26000000 I've noticed from the log file for all of my systems that the k0d and k0p values reach 1 really early (at about step 1210000) and then stay at 1 throughout. However, I'm afraid that it's a bit too fast. Is this behavior normal? Do my parameters look ok? If everything looks alright, is it ok that I shorten the equilibration time a bit, or should I keep it at 50 ns? Thank you so much for your help! All the best, Christine Qian _______________________________________________ GaMD-discuss mailing list GaM...@li...<mailto:GaM...@li...> https://lists.sourceforge.net/lists/listinfo/gamd-discuss [Screenshot 2025-06-23 at 13.39.31.png][Screenshot 2025-06-23 at 13.39.49.png] |
|
From: Miao, Y. <Yin...@me...> - 2025-06-20 21:22:59
|
Hi Christine, It seems ok to shorten your equilibration simulation. Meanwhile, you may set a larger number for “accelMDGStatWindow”, which could be roughly 4 times the total number of atoms in your system. 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 Editor-in-Chief npj Drug Discovery https://www.nature.com/npjdrugdiscov On Jun 20, 2025, at 1:52 PM, Christine Qian <chr...@uc...> wrote: You don't often get email from chr...@uc...<mailto:chr...@uc...>. Learn why this is important<https://aka.ms/LearnAboutSenderIdentification> Dear GaMD developers, I am running GaMD on several protein (a monomer of 42 residues)-membrane systems with waters and ions. The systems have sizes of about 170000 atoms. This is the parameters that I'm using for my equilibration: # GaMD Equilibration Params _________________________ accelMD on accelMDdual on accelMDdihe on accelMDG on accelMDGiE 1 accelMDGRestart off accelMDGcMDPrepSteps 200000 accelMDGEquiPrepSteps 200000 accelMDGStatWindow 20000 accelMDGcMDSteps 1000000 accelMDGEquiSteps 25000000 accelMDOutFreq 1000 accelMDGsigma0P 6.0 accelMDGsigma0D 6.0 run 26000000 I've noticed from the log file for all of my systems that the k0d and k0p values reach 1 really early (at about step 1210000) and then stay at 1 throughout. However, I'm afraid that it's a bit too fast. Is this behavior normal? Do my parameters look ok? If everything looks alright, is it ok that I shorten the equilibration time a bit, or should I keep it at 50 ns? Thank you so much for your help! All the best, Christine Qian _______________________________________________ GaMD-discuss mailing list GaM...@li...<mailto:GaM...@li...> https://lists.sourceforge.net/lists/listinfo/gamd-discuss |
|
From: Christine Q. <chr...@uc...> - 2025-06-20 18:08:08
|
Dear GaMD developers, I am running GaMD on several protein (a monomer of 42 residues)-membrane systems with waters and ions. The systems have sizes of about 170000 atoms. This is the parameters that I'm using for my equilibration: # GaMD Equilibration Params _________________________ accelMD on accelMDdual on accelMDdihe on accelMDG on accelMDGiE 1 accelMDGRestart off accelMDGcMDPrepSteps 200000 accelMDGEquiPrepSteps 200000 accelMDGStatWindow 20000 accelMDGcMDSteps 1000000 accelMDGEquiSteps 25000000 accelMDOutFreq 1000 accelMDGsigma0P 6.0 accelMDGsigma0D 6.0 run 26000000 I've noticed from the log file for all of my systems that the k0d and k0p values reach 1 really early (at about step 1210000) and then stay at 1 throughout. However, I'm afraid that it's a bit too fast. Is this behavior normal? Do my parameters look ok? If everything looks alright, is it ok that I shorten the equilibration time a bit, or should I keep it at 50 ns? Thank you so much for your help! All the best, Christine Qian |
|
From: Mehrnoosh K. H. <meh...@jh...> - 2025-06-11 15:12:54
|
Dear Dr. Yinglong, Thank you very much for your help. Yes, I am using hydrogen mass repartitioning for preparing my system. Best, Hazrati On Wednesday, June 11, 2025 15:43 CEST, "Miao, Yinglong" <Yin...@me...> wrote: Hi Hazrati, For your input parameters shown in the mdout file: &cntrl imin = 0, !No minimization irest = 1, !Restart the simulation, reading coordinates and velocities fromntx = 5, !Coordinates and velocities will be read from previous files nstlim = 1000500000, !Number of MD-steps to be performed """4000 ns""" dt = 0.004, tol = 0.000001, !The tolerance to use to determine when to stop the minimizatntc = 2, !bonds involving hydrogen are constrained ntf = 2, !to employ TIP3P set NTF = NTC = 2 ig = -1, !The seed for the pseudo-random number generator. the random seed wcut = 10.0, !non-bond cut off ntb = 2, !constant pressure ntp = 1, !isotropic position scaling ntpr = 250000, !Every ntpr steps, energy information will be printed in humantwx = 250000, !Every ntwx steps, the coordinates will be written to the mdcntwr = 250000, !Every ntwr steps during dynamics, the “restrt” file willntwprt = 0, !Include all atoms of the system when writing trajectories ntxo = 1, !Format of the final coordinate. 1 = ASCII, 2 = NetCDF ioutfm = 1, !Format of the coordinate and trajectory files. 1 = Binary NetCDiwrap = 1, !Wrap the molecule into the box. Good for long trajectories ntt = 3, !Thermostat type. 3 = Langevin dynamics gamma_ln = 2.0, !Collision frequency in ps-1 temp0 = 300.0, !Initial Temperature tempi = 300.0, !Aimed T ifsc = 1, !Soft core potential icfe = 1, !The basic flag for free energy calculations gti_cpu_output = 0, gti_add_sc = 1, irest_gamd = 0, igamd = 17, iE = 1, !Set the threshold enrgy to 2=upper bound, 1=lower bound iEP = 1, !Set the threshold enrgy to 2=upper bound, 1=lower bound iED = 1, !Set the threshold enrgy to 2=upper bound, 1=lower bound ntcmdprep = 500000, !2 ns ntcmd = 5000000, !20 ns !ntcmd = 0 ntebprep = 5000000, !20 ns nteb = 1000000000 !4000 ns !nteb = 0 ntave = 250, !every 1 ps sigma0P = 2.0, sigma0D = 6.0, timask1 = ':741-768', !Specifies the atoms unique to V0 in ambmask format. Resscmask1 = ':741-768', !Soft core mask timask2 = '', !Specifies the atoms unique to V1 in ambmask format. scmask2 = '', !Soft core mask bgpro2atm=1, edpro2atm=5647, nmropt = 1, / &wt type='REST', / &wt type='END', / DISANG=dist.RST / I have highlighted those that may cause issues. First, ntave may be too small; it should be roughly 4 times the total number of atoms in your system. Second, you may want to output energies and frames more frequently so that you have ~0.5 - 1 million frames in total from your production simulation for accurate reweighting. Next, for dt=0.004, make sure you have hydrogen mass repartitioning in your prepared system. Otherwise, it’s safer to start with dt=2fs simulations. Finally, you may break such a long simulation into multiple runs, especially separating the equilibration and production. You can find more information about the parameter settings and tutorials at: https://www.med.unc.edu/pharm/miaolab/resources/gamd/manual/, https://www.med.unc.edu/pharm/miaolab/resources/gamd/tutorial/, https://www.med.unc.edu/pharm/miaolab/resources/gamdtutorials/ Hope these help,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 Editor-in-Chief npj Drug Discovery https://www.nature.com/npjdrugdiscov ======= For discussion on GaMD, please subscribe to mailing list: https://sourceforge.net/projects/gamd/lists/gamd-discuss, post emails to "gam...@li..." and help others as much as you can! ======= On Jun 10, 2025, at 9:28 AM, Mehrnoosh Khodam Hazrati <meh...@jh...> wrote: Dear Dr. Yinglong, Thank you for your reply. Sorry, the file is large, I'm sending the first lines of the mdout file. Please let me know if I am doing something wrong. Thanks for your time, Best, Hazrati On Tuesday, June 10, 2025 14:08 CEST, "Miao, Yinglong" <Yin...@me...> wrote: HI Hazrati, Could you possibly share a copy of your mdout files, especially for the PPI-GaMD equilibration part? We can look into the issue more … 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 Editor-in-Chiefnpj Drug Discoveryhttps://www.nature.com/npjdrugdiscov On Jun 9, 2025, at 9:51 AM, Mehrnoosh Khodam Hazrati via GaMD-discuss <gam...@li...> wrote: Dear GaMD developer, Although I am using Amber22 for my simulations with PPI-GaMD-Dual, I get sharp changes in the dihedral energies, no matter if irest=1 or irest=0, and it leads to unexpectedly large Boost-Energy-Dihedral (the 8th column) in my gamd.log files. Since you mentioned in the manual that this bug is already solved in Amber20, I don't expect to face it using Amber22. I get reasonable values for the Boost-Energy-Potential (the 7th column). Do you have any idea if it is the bug or there might be something wrong with my setting? Thanks for your time. Best regards, Hazrati --Mehrnoosh Hazrati, Ph.D. | Czech Academy of SciencesAssociate ScientistE-mail: meh...@jh... _______________________________________________ GaMD-discuss mailing list GaM...@li... https://lists.sourceforge.net/lists/listinfo/gamd-discuss -- --Mehrnoosh Hazrati, Ph.D. | Czech Academy of SciencesAssociate ScientistE-mail: meh...@jh... |
|
From: Miao, Y. <Yin...@me...> - 2025-06-10 15:16:34
|
HI Hazrati, Could you possibly share a copy of your mdout files, especially for the PPI-GaMD equilibration part? We can look into the issue more … 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 Editor-in-Chief npj Drug Discovery https://www.nature.com/npjdrugdiscov On Jun 9, 2025, at 9:51 AM, Mehrnoosh Khodam Hazrati via GaMD-discuss <gam...@li...> wrote: Dear GaMD developer, Although I am using Amber22 for my simulations with PPI-GaMD-Dual, I get sharp changes in the dihedral energies, no matter if irest=1 or irest=0, and it leads to unexpectedly large Boost-Energy-Dihedral (the 8th column) in my gamd.log files. Since you mentioned in the manual that this bug is already solved in Amber20, I don't expect to face it using Amber22. I get reasonable values for the Boost-Energy-Potential (the 7th column). Do you have any idea if it is the bug or there might be something wrong with my setting? Thanks for your time. Best regards, Hazrati -- Mehrnoosh Hazrati, Ph.D. | Czech Academy of Sciences Associate Scientist E-mail: meh...@jh...<mailto:meh...@jh...> _______________________________________________ GaMD-discuss mailing list GaM...@li... https://lists.sourceforge.net/lists/listinfo/gamd-discuss |
|
From: Mehrnoosh K. H. <meh...@jh...> - 2025-06-09 14:08:58
|
Dear GaMD developer, Although I am using Amber22 for my simulations with PPI-GaMD-Dual, I get sharp changes in the dihedral energies, no matter if irest=1 or irest=0, and it leads to unexpectedly large Boost-Energy-Dihedral (the 8th column) in my gamd.log files. Since you mentioned in the manual that this bug is already solved in Amber20, I don't expect to face it using Amber22. I get reasonable values for the Boost-Energy-Potential (the 7th column). Do you have any idea if it is the bug or there might be something wrong with my setting? Thanks for your time. Best regards, Hazrati --Mehrnoosh Hazrati, Ph.D. | Czech Academy of SciencesAssociate ScientistE-mail: meh...@jh... |
|
From: Marc C. J. <Mar...@ua...> - 2025-06-05 08:32:01
|
Hi Jinan, Thank you for your insights, I will continue testing with different parameters and doing longer simulations with LiGaMD3, this time starting with a completely unbound ligand. With igamd=11 though, then it cannot support semi-isotropic coupling of pressure and with a membrane system, for long simulations it’s important. Thanks again for your comments, Marc Ciruela From: 王进安 <jin...@ex...> Date: Wednesday, 4 June 2025 at 07:14 To: Miao, Yinglong <Yin...@me...> Cc: Marc Ciruela Jardí <Mar...@ua...>, gamd-discuss <gam...@li...>, Joshi, Keya <ke...@em...>, Adediwura, Victor Ayo <vad...@em...>, Xiao, Fu <Fu...@me...> Subject: Re: Re: [GaMD-discuss] LiGaMD3 parameters and reweightening No soleu rebre correu electrònic de jin...@ex.... descobriu per què aquest fet és important<https://aka.ms/LearnAboutSenderIdentification> Hi Marc, I'll try my best. I have been trying to run GaMD on a membrane protein system with a small molecule and found several issues I would appreciate your help with. First, since I am trying to characterize the binding of the small molecule to the protein, should I start with the ligand bound to the protein or unbound? I don’t have a completely defined binding mode still, so it would be an approximation. The ligand binds in the interface of the protein with the membrane lipids, so in case I put it unbound, should be already in the membrane probably… For the initial conformation, I think it is fine to put it unbound. Another doubt I have is what kind of biasing to apply. LiGaMD3 seems to fit the best, but because the membrane is also somehow involved in the binding, should I use another kind of GaMD? Maybe the dual boost GaMD where the boost is applied to all the system? For identifying the ligand binding pose, LiGaMD3 should be the best one. But you can also try the mode of igamd=11 and 15. I tried applying LiGaMD3 with the threshold in the upper bound (iEP=2, iED=2 but iEB=1) as indicated in the LiGaMD3 publication and with a series of sigma0P and sigma0D values, from 4 to 8 with sigma0B fixed at 6. To decide what values to use, I looked at the trajectories and observed that with sigma0P=sigma0D=4 I already found an unbinding event. Looking at the log, k0P already stabilized at 1 and k0D and k0B were near 1. The values of the last unboosted potential energy and unboosted dihedral energy where 30.238509140909 and -226779.050457056845. The dihedral energy is very small, right? Why may this be happening? I continued with a production run but never found a rebinding event in 1us of simulation. As mentioned in the paper, in order to capture both binding and dissociation, we should choose the parameter of sigma0P and sigma0D to the value of which the largest sigma0P to keep the ligand in the bound state after the equilibration and at the same time, keeping the iEP and iED in the mode of 2 in the final of the equilibration. The values is the first and second boost potential, so 30.23 and -226779 seems good to me. Despite this, I decided to try and analyse the simulation to calculate the PMF. The dV distribution was centered around 30-25 kcal, which seems quite high, but I suppose it’s because of iE=2. The distribution seemed normal, but a bit skewed. I tried doing a reweightening using several reaction coordinates but the profiles where very sharp, I suppose because of the lack of re-binding processes which gives a bad sampling. The boost potential of 25-30 kcal/mol is quite good as the selective GaMD often generate higher boost. For the sharp PMF, yes, it is the bad sampling issue. With the very low energetic values of the dihedral energy and high dV, I decided to try iE=1 but then, with the same sigma values I did not obtain any unbinding during the equilibration and the values for the potentials remained similar during the equilibration. That's normal as the iE=1 put much lower boost in the system. Finally I tried with igamd=3 setting iEP=iED=1 and sigma0P=sigma0D=6. Here, I also didn’t see any unbinding but interestingly, the last unboosted potential energy and unboosted dihedral energy in the equilibration where -184058.981408115011 17220.120062542148, so completely different. The energy here is the total and dihedral potential. So it is different in different mode. Previously I had run a GaMD with a similar system with a peptide instead of a small molecule, using Pep-GaMD and starting from a quasi-bound state I was able to find a stable binding mode for the peptide. So I would much appreaicate your insight on what may be happening to this system and what recommendations would you have. For this one, you might test the same Pep-GaMD. Best, Jinan -----原始邮件----- 发件人:"Miao, Yinglong" <Yin...@me...> 发送时间:2025-06-04 04:57:32 (星期三) 收件人: "Marc Ciruela Jardí" <Mar...@ua...> 抄送: gamd-discuss <gam...@li...>, "Jinan Wang" <jin...@ex...>, "Joshi, Keya" <ke...@em...>, "Adediwura, Victor Ayo" <vad...@em...>, "Xiao, Fu" <Fu...@me...> 主题: Re: [GaMD-discuss] LiGaMD3 parameters and reweightening Dear Marc, Thanks for your great questions. I’m CC’ing a number of LiGaMD3 experts and hope they will be able to share their insights for the discussion. Best regards, 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 Jun 2, 2025, at 4:52 AM, Marc Ciruela Jardí via GaMD-discuss <gam...@li...> wrote: Dear GAMD developers, I have been trying to run GaMD on a membrane protein system with a small molecule and found several issues I would appreciate your help with. First, since I am trying to characterize the binding of the small molecule to the protein, should I start with the ligand bound to the protein or unbound? I don’t have a completely defined binding mode still, so it would be an approximation. The ligand binds in the interface of the protein with the membrane lipids, so in case I put it unbound, should be already in the membrane probably… Another doubt I have is what kind of biasing to apply. LiGaMD3 seems to fit the best, but because the membrane is also somehow involved in the binding, should I use another kind of GaMD? Maybe the dual boost GaMD where the boost is applied to all the system? In addition, I did some tests and have some doubts about the results. Following the indications in the tutorials, I did a GaMD equilibration (with a previously equilibrated system) with ntcmd = 1996360, ntcmdprep = 798544, nteb = 7985440, ntebprep = 798544 and ntave = 399272. I tried applying LiGaMD3 with the threshold in the upper bound (iEP=2, iED=2 but iEB=1) as indicated in the LiGaMD3 publication and with a series of sigma0P and sigma0D values, from 4 to 8 with sigma0B fixed at 6. To decide what values to use, I looked at the trajectories and observed that with sigma0P=sigma0D=4 I already found an unbinding event. Looking at the log, k0P already stabilized at 1 and k0D and k0B were near 1. The values of the last unboosted potential energy and unboosted dihedral energy where 30.238509140909 and -226779.050457056845. The dihedral energy is very small, right? Why may this be happening? I continued with a production run but never found a rebinding event in 1us of simulation. Despite this, I decided to try and analyse the simulation to calculate the PMF. The dV distribution was centered around 30-25 kcal, which seems quite high, but I suppose it’s because of iE=2. The distribution seemed normal, but a bit skewed. I tried doing a reweightening using several reaction coordinates but the profiles where very sharp, I suppose because of the lack of re-binding processes which gives a bad sampling. With the very low energetic values of the dihedral energy and high dV, I decided to try iE=1 but then, with the same sigma values I did not obtain any unbinding during the equilibration and the values for the potentials remained similar during the equilibration. Finally I tried with igamd=3 setting iEP=iED=1 and sigma0P=sigma0D=6. Here, I also didn’t see any unbinding but interestingly, the last unboosted potential energy and unboosted dihedral energy in the equilibration where -184058.981408115011 17220.120062542148, so completely different. Previously I had run a GaMD with a similar system with a peptide instead of a small molecule, using Pep-GaMD and starting from a quasi-bound state I was able to find a stable binding mode for the peptide. So I would much appreaicate your insight on what may be happening to this system and what recommendations would you have. Thank you for your time, Best regards, [cid:1cd9bdfb$1$1973954b49b$Coremail$jinan.wang$exeliris.ac.cn]<https://www.uab.cat/> Marc Ciruela Jardí Estudiant de doctorat Laboratori de Medicina Computacional Unitat de Bioestadística Edifici M - M3 · 08193 Bellaterra · Barcelona · Spain 93 581 38 12 www.uab.cat<https://www.uab.cat/> LMC<http://lmc.uab.cat/> [cid:5fc46ae$2$1973954b49b$Coremail$jinan.wang$exeliris.ac.cn]<https://orcid.org/0009-0001-9920-8789> Aquest missatge s'adreça exclusivament a la persona destinatària i pot contenir informació privada o confidencial. Si l'heu rebut per error, comuniqueu-nos-ho i destruïu-lo, i tingueu present que no teniu autorització per fer-ne cap ús. Aquest missatge no s’ha de respondre fora de la jornada laboral, en compliment de l’article 88 de la Llei orgànica 3/2018 de protecció de dades personals i garantia dels drets digitals<https://www.boe.es/boe_catalan/dias/2018/12/06/pdfs/BOE-A-2018-16673-C.pdf>, que reconeix el dret a la desconnexió digital, i de la Normativa interna per preservar el dret a la desconnexió digital del personal de la Universitat Autònoma de Barcelona<https://www.uab.cat/doc/NormativaDesconnexioDigital>. Abans d'imprimir aquest missatge penseu en el medi ambient. This message may contain private or confidential information, and is addressed exclusively to its intended recipient. If you have received it in error, please notify the sender and delete it. Please bear in mind that you are not authorized to use it for any purposes. Please consider the environment before printing this email. _______________________________________________ GaMD-discuss mailing list GaM...@li...<mailto:GaM...@li...> https://lists.sourceforge.net/lists/listinfo/gamd-discuss |
3 messages has been excluded from this view by a project administrator.
