Brokenportal,
Your questions are very important ones. We did not have a broad discussion of science behind FAH here in this forum, maybe we will move in that direction.
So are like, protein sequences randomly folding around in our computers then?
The process in only partly random. The movement of molecules of the solvent around the protein is random, but once part of a protein is moved in a way that it becomes locked with other parts, it doesn't move freely, even when it is being hit by the molecules of solvent. The forces between atoms of the protein is deterministic and so is the sequence of aminoacids. Therefore the result of this process is a protein that is folded in a cetrain determined way with certain probability of misfolding.
Then each time they reach a new fold, do they get analyzed for functionality? I mean, how can a computer tell what a particular computer generated protein fold is going to do in a given biological situation? Do you or anybody know?
When a protein folds in a "right" way (into native conformation, as biologists put this) it will have the same functionality no matter how long it folded and what were the transition states. It is like a jigsaw puzzle - no matter what the sequence of the pieces is during the assembly process, we're looking at the same picture every time it is completed.
When protein misfolds it can be either biologically active in a normal way (when there's only a small difference in conformaton, irrelevant to protein function) or it might become inactive. It can even get toxic - bilogically active in a harmful way - such as in prions - proteins causing mad cow's disease.
When a protein misfolds in FAH, it is not analyzed for functionality - the hundreds and thousands of interactions in the cell are (as of now) impossible to simulate.
Does anybody know if increased computing capacity can help sens more directly in any way like this?
Definitely, yes. We couldn't understand protein folding 10 years ago. 6-7 years ago FAH started to fold small proteins (such as villin), or parts of it. Currently FAH is running projects where there are over 100 000 atoms simulated.
FAH had 300 teraflops just 1.5 years ago, before PS3 client was launched. With PS3 it grew to 700 teraflops initially and to over 1000 teraflops (1 petaflop) some 10 months ago. The program broke the barrier of 2 petaflops this spring and with the launch of the GPU client it is currently churning out results at 2.5 petaflops. In the future the progress will likely be slower, but according to Moore's law we could see doubling of computational capacity every year or so.
I have tried to calculate what it would take to simulate a whole cell and, with some assumptions concerning increase in computational capacity and increase in number of participants, I came to a conclusion that a whole cell might be simulated before 2020. Smaller objects such as organelles (mitochondria, lysosomes, etc) or parts of them could be simulated couple of years earlier.
You are likely aware that one of the SENS strands, MitoSENS is dealing with the issue of importing 13 hydrophobic proteins from nucleus to the mitochondria. FAH could help in redesigning the proteins so that they behave better in the cytosol and could be imported through the mitochondrion membrane. Once FAH masters understanding of protein folding they will likely run some projects suggested or initiated by other parties that have the objective of designing therapies or understanding organelles functionality. It is likely that scientific institutions and foundations will outsource some part of their work to FAH, Methuselah Foundation included.
Maciek
Edited by naapi, 13 August 2008 - 07:45 AM.