Molecular Dynamics

44 views
Skip to first unread message

Josiah Zayner

unread,
Oct 30, 2012, 10:53:26 PM10/30/12
to diy...@googlegroups.com
Is anyone working on MD simulation stuff?
I know it can be a little complex starting out but it seems like an easy and great way for people in DIY Bio to contribute to science. Not just from computational power because there is lots of that but from a intellectual standpoint. Simulating stuff others don't or simulating stuff looking for interesting things.

Most simulation software is free and open including GROMACS http://www.gromacs.org and namd/CHARMM http://www.ks.uiuc.edu/Development/Download/download.cgi?PackageName=VMD


Bryan Bishop

unread,
Oct 30, 2012, 10:57:12 PM10/30/12
to diybio, Bryan Bishop, Josiah Zayner
On Tue, Oct 30, 2012 at 9:53 PM, Josiah Zayner <josiah...@gmail.com> wrote:
I know it can be a little complex starting out but it seems like an easy and great way for people in DIY Bio to contribute to science. Not just from computational power because there is lots of that but from a intellectual standpoint. Simulating stuff others don't or simulating stuff looking for interesting things.

There's some molecular dynamics stuff here:

Josiah Zayner

unread,
Oct 30, 2012, 11:25:56 PM10/30/12
to diy...@googlegroups.com, Bryan Bishop, Josiah Zayner
I was looking at that earlier today but there is not much detailed technical information on it the program.
The screen shot of the simulation tab doesn't even allow someone to change parameters. Temperature? What thermostat are you using?
When you run those simulations is there water? Vacuum? GROMACS 3.3 is very outdated especially if you are using the gmx forcefield. 
What simulation parameters? NVT? NPT? What is your electrostatics cutoff? Are you using PME? 
Are you running solvent restrained dynamics before your production run?
The pictures look great and all but it appears you are only applying forces to atoms to make them move that way? Or am I missing something?
The constructions also don't appear to be chemically relevant i.e. you just place random atoms next to each other without any care for charge or interaction.
Yellow is usually used to represent sulfur, why is it so heavily used in the designs? 
How long are the simulations in nanoseconds? 

Josiah Zayner

unread,
Oct 30, 2012, 11:29:57 PM10/30/12
to diy...@googlegroups.com, Bryan Bishop, Josiah Zayner
Interesting program though and cool idea. Current usability just seems limited to making cool gifs.

Nathan McCorkle

unread,
Oct 31, 2012, 12:08:26 AM10/31/12
to diy...@googlegroups.com

Josiah, you know a lot of lingo, have you used this stuff at all ever?

--
-- You received this message because you are subscribed to the Google Groups DIYbio group. To post to this group, send email to diy...@googlegroups.com. To unsubscribe from this group, send email to diybio+un...@googlegroups.com. For more options, visit this group at https://groups.google.com/d/forum/diybio?hl=en
Learn more at www.diybio.org
---
You received this message because you are subscribed to the Google Groups "DIYbio" group.
To post to this group, send email to diy...@googlegroups.com.
To unsubscribe from this group, send email to diybio+un...@googlegroups.com.
Visit this group at http://groups.google.com/group/diybio?hl=en.
To view this discussion on the web visit https://groups.google.com/d/msg/diybio/-/Zn5MpAi17pEJ.
For more options, visit https://groups.google.com/groups/opt_out.
 
 

rwst

unread,
Oct 31, 2012, 3:55:20 AM10/31/12
to diy...@googlegroups.com
I have tested NAMD with my NVIDIA and CUDA, very impressive. Now I would like to do protein folding with that setup,
applying forces from coupling analysis and a simulated annealing protocol. That's the idea, anyway. Did you know that
decades of research on human hemoglobine turned out to be wrong and you can do it right by patient application of MD?
There are so many interesting things to visualize, receptor docking, enzyme allosterisms, somebody stop me...

rwst

unread,
Oct 31, 2012, 4:12:38 AM10/31/12
to diy...@googlegroups.com
He's right. While you can see some (quite minor) Brownian motion the molecules were clearly simulated
in vacuum. They also appear unnaturally rigid. And the motion is arbitrarily applied, you can't see where
it comes from or why. Toys.

Josiah Zayner

unread,
Oct 31, 2012, 9:44:00 AM10/31/12
to diy...@googlegroups.com
Have you read the DE shaw 2010, 2011 science papers on folding? What are you looking at folding? 
Do you receive a huge speed-up from CUDA? 

I was doing some basic MD. I wanted to compare simulations of proteins that start from X-ray structures or NMR structures. One would predict that because of the nature of crystals that some X-ray structures might simulated differently due to starting in unnatural conformation. However, people think that X-ray crystallography is the gold standard in structural biology, predicting structure &c. when some of these structures are probably slightly, majorly off. I need to redo it though actually using NMR and X-ray structures from the same proteins instead of different proteins.

I was running on a Quad core but it might be good to just invest in something I can run CUDA with.
On a small protein <100 amino acids in a water box I could run maybe 1.5 nanoseconds a day.

Nathan McCorkle

unread,
Oct 31, 2012, 12:38:51 PM10/31/12
to diy...@googlegroups.com
On Wed, Oct 31, 2012 at 6:44 AM, Josiah Zayner <josiah...@gmail.com> wrote:
> I was doing some basic MD. I wanted to compare simulations of proteins that
> start from X-ray structures or NMR structures. One would predict that
> because of the nature of crystals that some X-ray structures might simulated
> differently due to starting in unnatural conformation. However, people think
> that X-ray crystallography is the gold standard in structural biology,
> predicting structure &c. when some of these structures are probably
> slightly, majorly off. I need to redo it though actually using NMR and X-ray
> structures from the same proteins instead of different proteins.

Are you saying that sometimes the way proteins crystallize isn't how
they are in situ, and you want to compare many different
crystallization methods with the structures produced to find your way
back to what they're really like in situ? Which you can then compare
with simulations?

Josiah Zayner

unread,
Oct 31, 2012, 12:51:30 PM10/31/12
to diy...@googlegroups.com
Protein crystallization has long been known to cause unnatural conformations in proteins due to the proteins been forced to adopt a highly dense rigid body interaction with other proteins in absence of bulk solvent. By unnatural I mean the conformations could perhaps be on the tail end of a Boltzmann distribution or have little representation in the solution ensemble. This is probably less often the case and the effects are usually more subtle. i.e. crystal contacts causing an alpha helix in the protein to adopt a slightly different 3D positioning than would actually occur.

NMR spectrscopy calculates structures using distance constraints and MD methods in solution to find highly represented conformers in the solution ensemble.
Proteins are inherently dynamic and access a wide range of conformations at room temperature in solution but they are still biasedly viewed as static structures. (See work by Dorothy Kern or Lewis Kay)

Comparing the RMSD progression of a structure over time during an MD simulation allows one to find the equilibrium or best fit relaxed states. Do these differ when starting from an X-ray or NMR and if so how.

rwst

unread,
Nov 1, 2012, 3:08:43 AM11/1/12
to diy...@googlegroups.com

>Have you read the DE shaw 2010, 2011 science papers on folding?
Thanks for the pointer, highly interesting! This might complement what I'm thinking of.


>What are you looking at folding?
I would like to help with unknown structures of M.tuberculosis proteins, but participating
at CASP would be exciting too. I have other duties on the table atm, though.


Do you receive a huge speed-up from CUDA? 

My reading is that it's at least 5-6x vs. quad core.
 
On a small protein <100 amino acids in a water box I could run maybe 1.5 nanoseconds a day.
 
You can probably speed that up by using a water sphere.
Ubiquitin in a water sphere (6,682 atoms) needs 16.12 seconds for 5 ps (2 fs steps),
that would be 26.8 nanoseconds/day on an AMD Phenom (6-core) and NVIDIA GTS 450.
You can get much better graphic cards now, however, the Tesla for example.

Try it yourself:
http://www.ks.uiuc.edu/Training/Tutorials/namd/namd-tutorial-unix-html/node8.html

rwst

unread,
Nov 1, 2012, 3:25:56 AM11/1/12
to diy...@googlegroups.com
Are you saying that sometimes the way proteins crystallize isn't how
they are in situ,

We must face this possibility. For example, many hemoglobine PDBs
turn out to be of improbable or unimportant conformation, and they have
to rewrite the book on how oxygen storage works with hemoglobine.
50 years of research down the drain!

T. Yonetani, M. Laberge: ''Protein dynamics explain the allosteric behaviors of hemoglobin.'' In: ''Biochimica et biophysica acta.'' 1784, 9, September 2008, 1146–1158. doi:10.1016/j.bbapap.2008.04.025. PMID 18519045. PMC 2668241. (Review).

Kanaori K, Tajiri Y, Tsuneshige A, Ishigami I, Ogura T, Tajima K, Neya S, Yonetani T. "T-quaternary structure of oxy human adult hemoglobin in the presence of two allosteric effectors, L35 and IHP." Biochim Biophys Acta. 2011 Oct;1807(10):1253-61. Epub 2011 Jun 15. PMID 21703224

Citation from abstract:
Therefore, the widely held assumptions of the structure-function correlation of Hb - [the deoxy-state]=[the T-quaternary structure]=[the low O(2)-affinity state] and [the oxy-state]=[the R-quaternary structure]=[the high O(2)-affinity state] and the O(2)-affiny of Hb being regulated by the T/R-quaternary structural transition - are no longer sustainable.
Reply all
Reply to author
Forward
0 new messages