- Posts: 70
problems with martini water in a TI calculation
- atsmonraz
-
Topic Author
- Visitor
7 years 10 months ago #5648
by atsmonraz
problems with martini water in a TI calculation was created by atsmonraz
Hi everyone, I've been working for a while now on parameterizing some lipidated aminoacids such as cysteine-palmitoyl vs all atom simulation values for the water-octanol partitoning coefficient. I've been working on the dh/dl vs lambda curves and while I'm getting relatively smooth curves for wet-octanol in martini as well as for all my all-atom simulations, I'm troubled by the curves i'm getting for martini water hydration TI which are very jagged. I'm using the recommended mdps that are published in this website to run martini via gromacs 5 with some modifications that i introduced to allow the free energy calculations so I'm fairly certain that's not the problem (although if anyone here thinks I should put the full mdps i'm using in the post, i have no problem to add them). Originally, I thought that I was having an undersampling problem and gradually went from 21 lambdas all the way up to 51 lambdas (with 2 us/lambda) and this is after a 100 ns NVT equilibration and another 100ns of NPT equilibration. In comparison to previously published papers that did similar works, it seems to me like an absurdly excessive protocol, yet i guess its still an option that it might be undersampled (at least theoretically). So at this point I'm just not sure what it is anymore...has anyone here previously encountered these kind of problems?
Most desperate for help,
Yoav
Most desperate for help,
Yoav
Please Log in or Create an account to join the conversation.
- jaakko
-
- Offline
- Admin
Less
More
7 years 10 months ago #5651
by jaakko
Replied by jaakko on topic problems with martini water in a TI calculation
Hi Yoav,
51 lambda points and 2us per window doesn't sound like an undersampling problem. Could you show the curves for a CG system where this problem occurs (water) and another where it doesn't (e.g. octanol)?
- Jaakko
51 lambda points and 2us per window doesn't sound like an undersampling problem. Could you show the curves for a CG system where this problem occurs (water) and another where it doesn't (e.g. octanol)?
- Jaakko
Please Log in or Create an account to join the conversation.
- atsmonraz
-
Topic Author
- Visitor
7 years 10 months ago - 7 years 10 months ago #5655
by atsmonraz
Replied by atsmonraz on topic problems with martini water in a TI calculation
Hi Jaakko,
Thanks for replying!
Here is the link
drive.google.com/file/d/0B3rWRyOz76rSdGg...aWM/view?usp=sharing
The upper one is martini water and the lower is martini wet octanol.
EDIT: red smooth line is the median values for the all-atom simulations measurements with the yellow colored region being their uncertainty and the blue line is the median of the martini simulations with the cyan colored region being its uncertainty. the turquoise colored area is the AA/CG uncertainty overlap.
These graphs were made after removing about ~5 points from the original graph at the lambda=0.3 and 0.7 region were their were sharp spikes that transitioned from the positive to the negative value of dh/dl (or vice versa) that made no sense at all. I can also post the original results if you're interested. Interestingly, I didn't have such spikes in the simulations that didn't use martini water.
Thanks for replying!
Here is the link
drive.google.com/file/d/0B3rWRyOz76rSdGg...aWM/view?usp=sharing
The upper one is martini water and the lower is martini wet octanol.
EDIT: red smooth line is the median values for the all-atom simulations measurements with the yellow colored region being their uncertainty and the blue line is the median of the martini simulations with the cyan colored region being its uncertainty. the turquoise colored area is the AA/CG uncertainty overlap.
These graphs were made after removing about ~5 points from the original graph at the lambda=0.3 and 0.7 region were their were sharp spikes that transitioned from the positive to the negative value of dh/dl (or vice versa) that made no sense at all. I can also post the original results if you're interested. Interestingly, I didn't have such spikes in the simulations that didn't use martini water.
Last edit: 7 years 10 months ago by atsmonraz. Reason: necessary details for the plots
Please Log in or Create an account to join the conversation.
- siewert
-
- Offline
- Admin
Less
More
- Posts: 61
7 years 10 months ago #5657
by siewert
Replied by siewert on topic problems with martini water in a TI calculation
Did you check whether the water box perhaps froze?
Please Log in or Create an account to join the conversation.
- atsmonraz
-
Topic Author
- Visitor
7 years 10 months ago - 7 years 10 months ago #5665
by atsmonraz
Replied by atsmonraz on topic problems with martini water in a TI calculation
Hi Prof. Marrink, thank you for your interest in this post, I appreciate it. As for your question - I ran all simulatons in 300K so I didn't think that the WF particles would be necessary. However, I did try to run one of the simulations with them just to be sure this isn't the case and still seen similar anomalies with the actual free energy difference being around about 2-3 kj/mol difference compared to the same simulation without WF particles. I tried using a 1:10 ratio in term of WF:W particles.
Last edit: 7 years 10 months ago by atsmonraz.
Please Log in or Create an account to join the conversation.
- jaakko
-
- Offline
- Admin
Less
More
- Posts: 70
7 years 10 months ago #5685
by jaakko
Replied by jaakko on topic problems with martini water in a TI calculation
Hi Yoav,
At 300K anti-freeze particles are already good to have always. Especially for free energy simulations that tend to have small boxes. Did you mean that you ran one of the windows using WF or one whole TI curve? If the latter could you show that one as well?
Have you compared the TI curve to what you'd get using BAR to calculate the free energy?
- Jaakko
At 300K anti-freeze particles are already good to have always. Especially for free energy simulations that tend to have small boxes. Did you mean that you ran one of the windows using WF or one whole TI curve? If the latter could you show that one as well?
Have you compared the TI curve to what you'd get using BAR to calculate the free energy?
- Jaakko
Please Log in or Create an account to join the conversation.
- atsmonraz
-
Topic Author
- Visitor
7 years 10 months ago - 7 years 10 months ago #5689
by atsmonraz
Replied by atsmonraz on topic problems with martini water in a TI calculation
Hi Jaakko, what I meant was that I ran an entire set of lambdas with WF particles but I did that back when I was trying 20-25 lambdas instead of 51. So just to make sure I'm covering all options I did a new run of 51 lambdas with WF particles over the last few days . Currently, I'm computing the results and as soon as I will have a curve ready i'll share it here.
You can see the old curve (24 lambdas with WF particles) I got in this link:
drive.google.com/file/d/0B3rWRyOz76rSaUt...aWM/view?usp=sharing
Interestingly, I tried constructing the figure with my own code and seem to get a much more continuous curve then I did previously (when I used alchemical gromacs). Must admit that it seems somewhat strange...
In anycase. I'll post the updated figure 51 lambdas later today.
EDIT: Jaakko attached is the new picture and unfortunately, I just noticed that I forgot to type the WF particles in the topology...so I guess you can disregard this plot. I'm now rerunning it again and this time I confirmed that the WF particles are in the system. Still, just in case it interests anyone here:
drive.google.com/file/d/0B3rWRyOz76rSQW1...eG8/view?usp=sharing
You can see the old curve (24 lambdas with WF particles) I got in this link:
drive.google.com/file/d/0B3rWRyOz76rSaUt...aWM/view?usp=sharing
Interestingly, I tried constructing the figure with my own code and seem to get a much more continuous curve then I did previously (when I used alchemical gromacs). Must admit that it seems somewhat strange...
In anycase. I'll post the updated figure 51 lambdas later today.
EDIT: Jaakko attached is the new picture and unfortunately, I just noticed that I forgot to type the WF particles in the topology...so I guess you can disregard this plot. I'm now rerunning it again and this time I confirmed that the WF particles are in the system. Still, just in case it interests anyone here:
drive.google.com/file/d/0B3rWRyOz76rSQW1...eG8/view?usp=sharing
Last edit: 7 years 10 months ago by atsmonraz. Reason: updated plot
Please Log in or Create an account to join the conversation.
- atsmonraz
-
Topic Author
- Visitor
7 years 10 months ago #5705
by atsmonraz
Replied by atsmonraz on topic problems with martini water in a TI calculation
Hey Prof. Marrink and Jaako, after a long 10 days I got my first proper TI calculation with water. Apparently in my previous attempt to gro file and topology weren't reading the WF particles properly and now they are after a whole bunch of issues that I had with the equilibration stages. But now it definitley works and the new plots are starting to come out and they're beautiful!
drive.google.com/file/d/0B3rWRyOz76rSRFR...T0U/view?usp=sharing
I can't express how much I appreciate your help, I've been at this for months on months now after trying everything and anything I could think of.
Much appreciated and all the best to you both !
drive.google.com/file/d/0B3rWRyOz76rSRFR...T0U/view?usp=sharing
I can't express how much I appreciate your help, I've been at this for months on months now after trying everything and anything I could think of.
Much appreciated and all the best to you both !
Please Log in or Create an account to join the conversation.
Time to create page: 0.103 seconds