Supercomputer Performs Simulation of Virus

moller writes to tell us Red Herring is reporting that researchers from the University of California at Irvine and the University of Illinois at Urbana-Champaign have announced that they created a computer simulation of a virus. From the article: "Using one of the world's fastest computers at the U.S. National Center for Supercomputing Applications, located at the University of Illinois at Urbana-Champaign, the researchers ran a computer program devised to reverse engineer the dynamics of all atoms making up the virus particle and a tiny drop of water containing it." Nature also has an interesting write up on the research surrounding this project.

These scientists must be Simpsons fans.

[Sebilrazen]

FTFA: This particular virus can only replicate in a cell which has already been infected by another virusthe tobacco mosaic viruswhich commonly attacks tomato plants.

Tomacco.

More/better information...

[tskirvin]

1. The full research page for this project is here [uiuc.edu]. This is a lot better than the stuff linked through Nature and such.

2. The image was actually generated by our group [uiuc.edu], and specifically Anton Arkhipov [uiuc.edu], using our software package VMD [uiuc.edu]. NCSA didn't have anything to do with it.

Re:You can run this yourself (theoretically)

[tskirvin]

Also of note: we've done a series of cluster-building workshops [uiuc.edu] specifically focusing on the software and hardware required to run these kinds of simulations. Copies of the presentations and tutorials are linked off of that page.

And if you want to see how we designed our clusters, I've got full specifications up here [uiuc.edu].

Re:This is why we do scientific computing

[Anonymous Coward]

The evidence exists and is well documented.

Research in this field is still fairly new, but there is a community of scientists that specialize on this exact topic (I'm one of them). For more well-established and respected individuals, look up the work by Brian Bothner, Gary Siuzdak, Jack Johnson, Adam Zlotnick, JK Lewis, Peter Prevelige, etc (apologies for leaving people out).

As a shameless plug, my current research involves writing software to link our experimental results to the computer models of motion.

Re:No, this is scientific showboating.

[Anonymous Coward]

There may not be a complete simulation of a protein folding, but there is plenty of useful information that can be discovered about how a folded protein moves about, how motions might make the active site available etcetera. Much of this information can be cross-checked and validated somewhat using other methods, such as NMR.

Ab-initiio protein folding is a much larger problem than using MD to find something out about a protein or polypeptide. Finding out more about the motions of an already folded protein can be really useful.

Re:No, this is scientific showboating.

[Anonymous Coward]

>>>You give them too much credit... force field people compute *some* of the forces and ignore most of them. Long-range electrostatics are often omitted entirely (and people wonder why their RNA strands fall apart once it flops around a bit - I mean jebus, people!).

Hey moron, do YOU work in the field?

Long range electrostatics are included in the form of either a cutoff function (usually around 14 A) or, much more commonly, a pme method which doesn't cutoff electrostatics but instead maps charges to a grid and calculates interactions through a fast fourier transform (something to that effect).

Re:No, this is scientific showboating.

[Salis]

http://www.stanford.edu/group/pandegroup/ [stanford.edu]

They've folded proteins whose kinetics are 1st order (ie. small enough proteins).

The folding problem is the one of the hardest ones. So don't get all blustery about it not being a solved problem. Cancer hasn't been cured yet either.

Re:No, this is scientific showboating.

[Decaff]

You're an idiot.

No, a scientist; and one who has worked in this area

Parameters are developed to reproduce very basic and general properties of molecules. Nothing is changed to go from one system to the next. So results that are reached are indeed unique and interesting and were certainly not built in to the system prior.

Nonsense. There are a range of parameter sets you use depending on the simulation you want to perform. For example, there are a wide range of models that are used for water alone! (such as SPC, TIP4 etc.) Each gives different results depending on the type of simulation, and things like the scale you are operating at. Then there are the different parameter sets you can choose to represent the protein surfaces.

If you want to do a large-scale model of a virus in water, you pick the parameter sets that you already know from similar simulations (large protein clusters in water etc.) that will give good results. So, of course it is going to be realistic.

A Poster Giving an Overview of the Whole Thing

[Expert Determination]

Here [uiuc.edu]. They're just looking at what holds together the structure of the virus.

Re:No, this is scientific showboating.

[Decaff]

In order to get the more "trustable" simulations to produce something in the ballpark of remotely representing reality, you have to know "the answer" before you do the simulation and then teach the model to reproduce that "answer"... then you can write a paper and show that you're model "get's the answer" - and that's about the limit of "insight" that's often gained from these sort of simulations.

You are, of course, absolutely right. Things haven't changed since I started to do this kind a long time (20 years) ago. However, you can gain some sort of insights, as you can find out which interactions (even thought they are simplified) can matter in the real world. By trying to simulate the real world, you can find out more about it, even if you aren't able to make many predictions.

Re:No, this is scientific showboating.

[AFairlyNormalPerson]

If your system is pure MM and described by point charges (no higher order multipole moments included, e.g. the TINKER force field), then particle mesh ewald will get you the interaction energy between the "real" unit cell and all possible images of that cell. There is still, of course, the real space interactions WITHIN the cell and these either use a cutoff function or some people use recursive bisection/fast mutipole methods (which is much better and can be applied to non-periodic systems -- Look for papers by Perez-Jorda. Er. His last name is hyphenated; it's not 2 people.).

If your MM model contains more than monopole functions, then you are "kind of" screwed (according to one of the developers of the TINKER package) -- Look for papers by Jay Ponder.. he has probably the most advanced MM force field out there. No "good" PME nor FMM scheme has been worked out to handle this.

If your system has a QM region within it, well, I've only known one person to use PME with that... because he's sitting next to me and was the person who implemented it into CHARMM and AMBER a couple of months ago :) (The QM atoms are just treated as points with the mulliken charge). -- Look for papers by Jaili Gao.