trouble with MC
-
- Posts:29
- Joined:Mon Feb 09, 2009 10:58 am [phpBB Debug] PHP Warning: in file [ROOT]/vendor/twig/twig/lib/Twig/Extension/Core.php on line 1275: count(): Parameter must be an array or an object that implements Countable
Hello everybody,
after running several MC simulations in hippo with a nearly perfect helical protein, I tried to interpret the results. As I expected, the protein in an normal GBSA formed a random coil. If you use this coil as a new input for a MC simulation with GBIM you can actually see, that the protein is moving and wobbling depending on the start position. Now there has been some unexpected effects in a slightly different simulation with the same protein:
I made a setup in which the nearly complete alpha helical protein was positioned near the interface region of the implicit membrane. So the protein starts as a long, helical poly-peptide and not as a random coil (because I start with the unsovlated pdb). After simulating for 1 million steps in an MC GBIM the movie.pdb shows very different movements for the protein. The protein seems to be moved as a whole unit and doesn't coil a bit (looks extremely artificial). I am very confused, because of the fact, that the protein isn't coiling even that it is in the same dielectric environment (it should coil like in an pure water GBSA simulation). So I tried some alternative MC simulations with different seed numbers and changed specific values, but nothing changed : The protein is still behaving like it would have a restrained backbone. I don't understand why a "pre-solvated" protein is behaving so diversely in comparison with an unsolvated one.
I hope you understand my problem and I'm looking forward for your hints and tips to improve and solve this case.
Thx a lot ,
your M.Kalavera
after running several MC simulations in hippo with a nearly perfect helical protein, I tried to interpret the results. As I expected, the protein in an normal GBSA formed a random coil. If you use this coil as a new input for a MC simulation with GBIM you can actually see, that the protein is moving and wobbling depending on the start position. Now there has been some unexpected effects in a slightly different simulation with the same protein:
I made a setup in which the nearly complete alpha helical protein was positioned near the interface region of the implicit membrane. So the protein starts as a long, helical poly-peptide and not as a random coil (because I start with the unsovlated pdb). After simulating for 1 million steps in an MC GBIM the movie.pdb shows very different movements for the protein. The protein seems to be moved as a whole unit and doesn't coil a bit (looks extremely artificial). I am very confused, because of the fact, that the protein isn't coiling even that it is in the same dielectric environment (it should coil like in an pure water GBSA simulation). So I tried some alternative MC simulations with different seed numbers and changed specific values, but nothing changed : The protein is still behaving like it would have a restrained backbone. I don't understand why a "pre-solvated" protein is behaving so diversely in comparison with an unsolvated one.
I hope you understand my problem and I'm looking forward for your hints and tips to improve and solve this case.
Thx a lot ,
your M.Kalavera
-
- Developer
- Posts:38
- Joined:Sun Jun 15, 2008 10:27 pm [phpBB Debug] PHP Warning: in file [ROOT]/vendor/twig/twig/lib/Twig/Extension/Core.php on line 1275: count(): Parameter must be an array or an object that implements Countable
Re: trouble with MC
Hi,
It could be that the rigid body moves dominate. These moves were put in to speed up the spatial orientation in the membrane but might be problematic at the start. If you want the protein to stay where it is and not perform any rigid body motion you can set the prob rigidbodymove/MC scan to zero. Then it will just do concerted rotation and sidechain moves. Have you checked the acceptance ratio of the MC moves? It should be in the output.dat file.
Let me know if this was the issue!
Good luck!![Smile :)](./images/smilies/icon_e_smile.gif)
Martin
It could be that the rigid body moves dominate. These moves were put in to speed up the spatial orientation in the membrane but might be problematic at the start. If you want the protein to stay where it is and not perform any rigid body motion you can set the prob rigidbodymove/MC scan to zero. Then it will just do concerted rotation and sidechain moves. Have you checked the acceptance ratio of the MC moves? It should be in the output.dat file.
Let me know if this was the issue!
Good luck!
![Smile :)](./images/smilies/icon_e_smile.gif)
Martin
-
- Posts:29
- Joined:Mon Feb 09, 2009 10:58 am [phpBB Debug] PHP Warning: in file [ROOT]/vendor/twig/twig/lib/Twig/Extension/Core.php on line 1275: count(): Parameter must be an array or an object that implements Countable
Re: trouble with MC
Hi Martin,
I tried the same simulation's with prob rigidbodymove/MC scan set to zero. As you promised, the protein seems to be bound to the starting position and only the sidechains and the backbone is moving. But exactly this is my problem. I just expect the protein to denature or form a random coil, as it does in pure water simulation (GBSA). For this purpose , I set up a GBIM/MC and moved the protein above the membrane into the water surrounding. In my imagination, the protein should perform the same strong movements as it did in the GBSA simulation, but performed simulations show a complete different result. The protein moves up and down and rotates, but seems to be stiff and restrained in the backbone, which looks totally artificial to me. I don't understand the non-coiling behavior of the protein in the MC/GBIM simulations in comparison to the GBSA simulation. I hope you understand my problem and appreciate your help. Maybe you can come up with new ideas,
your M.Kalavera
I tried the same simulation's with prob rigidbodymove/MC scan set to zero. As you promised, the protein seems to be bound to the starting position and only the sidechains and the backbone is moving. But exactly this is my problem. I just expect the protein to denature or form a random coil, as it does in pure water simulation (GBSA). For this purpose , I set up a GBIM/MC and moved the protein above the membrane into the water surrounding. In my imagination, the protein should perform the same strong movements as it did in the GBSA simulation, but performed simulations show a complete different result. The protein moves up and down and rotates, but seems to be stiff and restrained in the backbone, which looks totally artificial to me. I don't understand the non-coiling behavior of the protein in the MC/GBIM simulations in comparison to the GBSA simulation. I hope you understand my problem and appreciate your help. Maybe you can come up with new ideas,
your M.Kalavera
-
- Developer
- Posts:38
- Joined:Sun Jun 15, 2008 10:27 pm [phpBB Debug] PHP Warning: in file [ROOT]/vendor/twig/twig/lib/Twig/Extension/Core.php on line 1275: count(): Parameter must be an array or an object that implements Countable
Re: trouble with MC
Hi,
I just checked and the behavior is indeed correct. The reason for the slower sampling when using GBIM rather than GBSA is that the coulomb interactions are not compensated by solvation forces in the membrane. The membrane part is therefore comparable to a vacuum simulation. In normal GBSA the coulomb energy is anti-correlated to the GB energy, making the overall energy landscape smooth and thus aiding transitions. One way around this is to use replica exchange (REMC) or increase the temperature. For peptides its usually fine to go to 350K. Or use replica exchange 300-600K. 10 replicas are usually fine.
Hope this helps and good luck!
Best wishes,
Martin
I just checked and the behavior is indeed correct. The reason for the slower sampling when using GBIM rather than GBSA is that the coulomb interactions are not compensated by solvation forces in the membrane. The membrane part is therefore comparable to a vacuum simulation. In normal GBSA the coulomb energy is anti-correlated to the GB energy, making the overall energy landscape smooth and thus aiding transitions. One way around this is to use replica exchange (REMC) or increase the temperature. For peptides its usually fine to go to 350K. Or use replica exchange 300-600K. 10 replicas are usually fine.
Hope this helps and good luck!
Best wishes,
Martin
-
- Posts:29
- Joined:Mon Feb 09, 2009 10:58 am [phpBB Debug] PHP Warning: in file [ROOT]/vendor/twig/twig/lib/Twig/Extension/Core.php on line 1275: count(): Parameter must be an array or an object that implements Countable
Re: trouble with MC
Hi Martin,
I tried some new simulations and setups with the same system (GBIM MC and alpha helical protein) we have been talking above. At first, I changed the temperature to the value of 498 K and set the value of prob ridigbodymove to 0.1 .This change showed no effects on the protein behaving. It doesn't coil at all and moves with a stiff backbone at all. At a closer look the backbone is moving, but it seems more like relaxing (really small movements) than heavy movements, which I would expect at this temperature.
For the second simulation I chose 0.5 for the prob ridigbodymove value and to be absolutely sure to avoid positioning the protein directly into the membrane, I moved it additional about 100 angstroem in the z-direction. So the protein has been moved about 115 angstrom along the z-axis and the temperature is still set to 498 K.
After all the protein won't coil and is still a stiff alpha helix.
Are there some other setups or value changes I should try to get a different result ???
Thx for your support,
your M.Kalavera
I tried some new simulations and setups with the same system (GBIM MC and alpha helical protein) we have been talking above. At first, I changed the temperature to the value of 498 K and set the value of prob ridigbodymove to 0.1 .This change showed no effects on the protein behaving. It doesn't coil at all and moves with a stiff backbone at all. At a closer look the backbone is moving, but it seems more like relaxing (really small movements) than heavy movements, which I would expect at this temperature.
For the second simulation I chose 0.5 for the prob ridigbodymove value and to be absolutely sure to avoid positioning the protein directly into the membrane, I moved it additional about 100 angstroem in the z-direction. So the protein has been moved about 115 angstrom along the z-axis and the temperature is still set to 498 K.
After all the protein won't coil and is still a stiff alpha helix.
Are there some other setups or value changes I should try to get a different result ???
Thx for your support,
your M.Kalavera
-
- Developer
- Posts:38
- Joined:Sun Jun 15, 2008 10:27 pm [phpBB Debug] PHP Warning: in file [ROOT]/vendor/twig/twig/lib/Twig/Extension/Core.php on line 1275: count(): Parameter must be an array or an object that implements Countable
Re: trouble with MC
Hi,
Sorry you're still having problems.
If you attach your input file, (and maybe the movie.pdb) i'll have a look.
Best wishes,
Martin
Sorry you're still having problems.
If you attach your input file, (and maybe the movie.pdb) i'll have a look.
Best wishes,
Martin
-
- Posts:29
- Joined:Mon Feb 09, 2009 10:58 am [phpBB Debug] PHP Warning: in file [ROOT]/vendor/twig/twig/lib/Twig/Extension/Core.php on line 1275: count(): Parameter must be an array or an object that implements Countable
Re: trouble with MC
Here is one of the hippo_input.txt files I use.
They are all nearly the same except for the concrete values. There are no additional entries in one off the other input files.
hippo input file
mode MC simulation
input type [PDB,zmx] PDB
input file minimized.pdb
random number seed (must be < 0) -100
set_helix_backbone 0
use_pdb_coordinates 1
match_hydrogens 0
max_steps 1000000
T [K] 298.15
use_SSE (0 = slow double prec.) 1
use_pbc 0
use_cutoff 0
freq_pdb_out 5e2
freq_xyz_out 1e4
freq_E_out 1e2
freq_state_restart_out 1e5
print_speed 1
conrot window_move_prob 0.5
conrot a 150
fixed chain ends 0
prob rigidbodymove/MC scan 0.0
max rigidbody move size [A] 0.5
use_gbsa 1
gb_mode 2
gb_mode2_printdetails 0
gb_cutoff [A] 8
united atom for aliphatic C 0
r_probe [A] 1.4
sigma 0.005
intra molecule epsilon 2.0
use_gb_flex123 (flexible 1-2 1-3) 1
use_gb_membrane 1
sigma_membrane -0.030
born_mode [1=exp 2=gauss] 2
sasa_mode [0,1=v(z) 2=gauss 3=exp] 2
v(z)_mode [1=linear 2=gauss 3=exp] 2
gamma_born -2.0
gamma_born2 -2.0
gamma_sasa -1.5
gamma_v(z) -2.0
L [A] 15
openMP numthreads 1
I am really happy about your strong intense to help me,
thx allot for all your time and work.
Your M.Kalavera
They are all nearly the same except for the concrete values. There are no additional entries in one off the other input files.
hippo input file
mode MC simulation
input type [PDB,zmx] PDB
input file minimized.pdb
random number seed (must be < 0) -100
set_helix_backbone 0
use_pdb_coordinates 1
match_hydrogens 0
max_steps 1000000
T [K] 298.15
use_SSE (0 = slow double prec.) 1
use_pbc 0
use_cutoff 0
freq_pdb_out 5e2
freq_xyz_out 1e4
freq_E_out 1e2
freq_state_restart_out 1e5
print_speed 1
conrot window_move_prob 0.5
conrot a 150
fixed chain ends 0
prob rigidbodymove/MC scan 0.0
max rigidbody move size [A] 0.5
use_gbsa 1
gb_mode 2
gb_mode2_printdetails 0
gb_cutoff [A] 8
united atom for aliphatic C 0
r_probe [A] 1.4
sigma 0.005
intra molecule epsilon 2.0
use_gb_flex123 (flexible 1-2 1-3) 1
use_gb_membrane 1
sigma_membrane -0.030
born_mode [1=exp 2=gauss] 2
sasa_mode [0,1=v(z) 2=gauss 3=exp] 2
v(z)_mode [1=linear 2=gauss 3=exp] 2
gamma_born -2.0
gamma_born2 -2.0
gamma_sasa -1.5
gamma_v(z) -2.0
L [A] 15
openMP numthreads 1
I am really happy about your strong intense to help me,
thx allot for all your time and work.
Your M.Kalavera
-
- Developer
- Posts:38
- Joined:Sun Jun 15, 2008 10:27 pm [phpBB Debug] PHP Warning: in file [ROOT]/vendor/twig/twig/lib/Twig/Extension/Core.php on line 1275: count(): Parameter must be an array or an object that implements Countable
Re: trouble with MC
Hi,
Sorry for the delay!
The file looks fine. But your values must start at column number 41, otherwise it will not work. Hippo will print a complete input file with all the parameters you set into log.txt. Please check that it's fine. I think the behavior you see is correct. I'm currently running a folding simulation and it's usually stuck in a conformation for 20-100 million MC steps. This is typical in folding. For folding runs it is best to run >1e9 MC steps, and elevate the temperature as much as possible for your system. Or use REMC (replica exchange), that will massively speed up sampling.
Let me know what you think.
best wishes,
martin
Sorry for the delay!
The file looks fine. But your values must start at column number 41, otherwise it will not work. Hippo will print a complete input file with all the parameters you set into log.txt. Please check that it's fine. I think the behavior you see is correct. I'm currently running a folding simulation and it's usually stuck in a conformation for 20-100 million MC steps. This is typical in folding. For folding runs it is best to run >1e9 MC steps, and elevate the temperature as much as possible for your system. Or use REMC (replica exchange), that will massively speed up sampling.
Let me know what you think.
best wishes,
martin
-
- Posts:29
- Joined:Mon Feb 09, 2009 10:58 am [phpBB Debug] PHP Warning: in file [ROOT]/vendor/twig/twig/lib/Twig/Extension/Core.php on line 1275: count(): Parameter must be an array or an object that implements Countable
Re: trouble with MC
Hi Martin,
sorry for the bad formating of my values
but in the normal case the values start proper at column number 41, else wise I get the normal errors for a bad input in the log.txt.
What could be a problem is my lag of experience in MC simulations, but I thought that 10*10^6 would be a good start. I can try a run with the same setup for at least 1e9 steps , but I think this will take a while. I will reply, when its finished.
Thanks for your help ,
your M.Kalavera
sorry for the bad formating of my values
![Embarrassed :oops:](./images/smilies/icon_redface.gif)
What could be a problem is my lag of experience in MC simulations, but I thought that 10*10^6 would be a good start. I can try a run with the same setup for at least 1e9 steps , but I think this will take a while. I will reply, when its finished.
Thanks for your help ,
your M.Kalavera
-
- Posts:29
- Joined:Mon Feb 09, 2009 10:58 am [phpBB Debug] PHP Warning: in file [ROOT]/vendor/twig/twig/lib/Twig/Extension/Core.php on line 1275: count(): Parameter must be an array or an object that implements Countable
Re: trouble with MC
Hello Martin,
after a week of calculations, my MC simulation finished. I proceeded for about 500000000 MC-Steps. Even with this long period, the protein still keeps acting very rigid and stiff (optic control) and I am still wondering, if everything is alright. Only in the last Steps there are some movements in which the conformation is slightly changed. It looks quiet mysterious to me. Is there something else I could try or check ?
Your, M.Kalavera
after a week of calculations, my MC simulation finished. I proceeded for about 500000000 MC-Steps. Even with this long period, the protein still keeps acting very rigid and stiff (optic control) and I am still wondering, if everything is alright. Only in the last Steps there are some movements in which the conformation is slightly changed. It looks quiet mysterious to me. Is there something else I could try or check ?
Your, M.Kalavera