Opponents of animal testing for medical research often argue that the same tests could be performed via computer simulation; researchers counter that simulations simplify physiology too much to be useful in that way. But such a claim may be in its final era -- we now have the first functional, down-to-the-atom simulation of a biological organism. Computational biologists at the University of Illinois at Urbana-Champaign and crystallographers at the University of California at Irvine have created a complete simulation of the Satellite Tobacco Mosaic virus. We won't have a SimRabbit, SimRhesus Monkey or SimHuman any time soon, but such tools now appear to be much more plausible.
The satellite tobacco mosaic virus is about as simple a virus as possible; the entire STMV genome consists of a little over a thousand nucleotides in RNA, along with a couple of proteins. The virus is referred to as a "satellite" because it relies on the presence of the tobacco mosaic virus in order to reproduce. Despite this simplicity, the researchers had to use a supercomputer to simulate a fraction of a second of viral activity:
Running on a machine at the National Center for Supercomputing Applications, Urbana, the program calculated how each of the million or so atoms in the virus and a surrounding drop of salt water was interacting with almost every other atom every femtosecond, or millionth of a billionth of a second.
The team managed to model the entire virus in action for 50 billionths of a second. Such a task would take a desktop computer around 35 years, says Schulten. "This is just a first glimpse," he says. "But it looks gorgeous."
The researchers have a page describing the work in fairly complex language; Nature reduces the jargon a bit; and the University of Illinois-Urbana Champaign press release gives an essentially jargon-free depiction.
Virus simulations at this level, even without a corresponding simulation of a host organism, can reveal surprising details about viral activity; this first simulation included its own breakthrough discovery about how the virus creates its protein shell. As these tools become more advanced, we should expect to see similar discoveries of the subtle behavior of viruses, bacteria and beyond. Eventually, we'll likely be able to simulate the effects of changes to the organism's genome. This will be a major advance in our ability to predict the effects of bioengineering experiments, and to prepare for non-obvious results.
Comments (8)
Wow. This is amazing. I can't wait to see what happens in the coming decade with increasing computing power.
Posted by Benjamin Farahmand | March 14, 2006 4:16 PM
Posted on March 14, 2006 16:16
This is a benchmark. This atomic-level virus simulation requires a supercomputer now but, in another 5 years, another 10?
And the model may teach us shortcuts and optimizations to simplify future models without sacrificing accuracy.
Me, I'm waiting for the atomic-level neuron simulation.
Posted by Pace Arko | March 15, 2006 1:47 AM
Posted on March 15, 2006 01:47
Humberto Jesus
DeNAe
go out & play
autopoieisis self
determination cell clone
sim-gene
Posted by Glen Shorts | March 15, 2006 5:14 PM
Posted on March 15, 2006 17:14
What does this mean for computer viruses? Can we expect far more complex, biologically-inspired software viruses in the future?
Posted by Doug | March 15, 2006 5:53 PM
Posted on March 15, 2006 17:53
There is a shortcoming about this simulation that occurred to me earier today. What about the protein folding problem? We still don't have good models for the dynamics of protein molecules so how can we be sure this virus particle simulation is accurate?
I'm not saying this problem is insoluable. I'm just saying that atomic level accuracy is not there just yet.
Posted by Pace Arko | March 15, 2006 10:21 PM
Posted on March 15, 2006 22:21
Regarding the protein folding comment... the problem is that protein folding is thought to occur on the timescale of milliseconds (1e-3 sec) while with current computer technology we can only simulate in the 100s of nanoseconds range (1e-9). For example, this virus was only simulated for 10s of nanoseconds. So it's not that it isn't possible, just that we can't do it yet. Of course, some people have been able to fold small peptide, and there are simplifications to make it faster, but another problem is that nobody is sure of the exact mechanism of folding yet.
Also, the striking insight is more that the protein capsid is not stable without the RNA component that makes up the viral genome, not about how the virus creates its protein shell.
Posted by Naveen Michaud-Agrawal | March 16, 2006 6:26 AM
Posted on March 16, 2006 06:26
I don't think this method has any real chance of modelling something even on the cellular level. A virus is only one molecule.
Trying to model every atom in a cell would be (in my opinion) impossible, as the cell is constantly passing molecules back and forth with the body. Calculations on this huge scale would certainly be limited by the fact that it would be a highly chaotic system, and most like transcomputational.
You could assume every atom is a variable in 10 or more dimensions, and that there are trillions of atoms in a cell and trillions of cells in a body.
By the way, I found this take on the story much more interesting than what I read on Slashdot.
Posted by Ben Wendt | March 16, 2006 6:50 AM
Posted on March 16, 2006 06:50
Ben writes, "Trying to model every atom in a cell would be (in my opinion) impossible...."
Enormously difficult, yes, but let us hesitate to say impossible. Maybe we could do it with quantum computing. Maybe some revolution in mathematics will allow us to enormously simplify the equations that govern the dynamics. I realize that's handwaving but impossible is such an absolute word.
Regardless, this simulation might still teach us things even if it simplifies the details.
Posted by Pace Arko | March 16, 2006 7:25 AM
Posted on March 16, 2006 07:25