Hi
Yes i would tend to agree. Although we live in the early days of applied
super computing and they are at risk of technical embarrassment if
someone gets real specific with good math and ends up crunching at
effective rates that are several time the magnitude of the ibm
methodology. An example
that cane to me was at a recent geophysics convention i saw a super
cluster (7 or 8 tera flop ) that could not produce a solution to the
wave equation in the same time frame as a costom built device that i
built three yrs ago that only runs a rate 250 giga flops(equivelent).
The difference was system efficiency.
Your are right about the evolution of the math . if there not certain
then they have to be very general and hence safe.
Tim
Artem Evdokimov wrote:
>> IBM people are good engineers. However, my suspicion is that they would
> build the machine and try to invent protein folding algorhythms at the same
> time. The machine can be designed using general principles of how the
> protein folding routine would work, but without actually ironing out all the
> details before beginning. IMHO, this is a giant advertisement for IBM, just
> as Deep blue was :)
>> A.G.E.
>> "Tim Davies" <daviest at shaw.ca> wrote in message
> news:3C352393.F39DB0B at shaw.ca...> > i just seem to remember that the folks who sort of did the human dna
> > thing last year have a jv with ibm to build a box of around 30tera flop
> > or so. It was to do structure of proteins, so naturally they believe
> > that crunch power is the key to some aspect of the problem. There box
> > will be only partially efficient , so i was thinking that they must be
> > very very sure of there math to build a 100mill box that is only a super
> > cluster type device. Where there is confidence in the math , there are
> > much more efficient routs than the method chosen. so i wonder is the
> > math known , and what is it. does it fit a fast methodology . The
> > problem , from a pure math perspective , must have been structured and
> > tested prior to the build , or someone has more money than brains.
> > maybe
> > tim
> >
> > "D.K." wrote:
> > >
> > > Tim Davies <daviest at shaw.ca> wrote:
> > > >My thanks for the reply:
> > > >The problem would seem to be the computational limitations. In both of
> > > >the cases that you feel are satisfactory ( the qm and orbital (which
> > > >sound similar)) have very large computational requirements. This is
> > > >similar to the solution to the elastic wave equation in 10 or 10 k
> > > >velocity layers in the case of sound (it cant be done yet in its
> > > >complete form). I cant help but wonder if you did have a box that would
> > > >do 100 or 200 trillion instructions per sec then would it be possible?.
> > > >only if there is something to model can this be determined. I think its
> > > >a problem worth looking into. There seems to be a lot of interest in
> the
> > > >resulting structures. Im not sure why there is interest other than it
> > > >saves time, but the computational complexity is most interesting to me
> > > >Thanks
> > > >Tim Davies
> > >
> > > I am not an expert in the field, but... I doubt processing power alone
> > > is a limitation. As Artem noted, folding of proteins is aided by
> > > other proteins and such interactions are poorly understood. Also, at
> > > least some folds, if not most, appear to fold not into global free
> energy
> > > minimum. I worked with several proteins that fold spontaneously in
> > > solution into something that has seemingly nothing to do with their
> > > native folds.
> > >
> > > As for computational power involved, I vaguely remember an article
> > > back from few years ago that illustrates the problem well. There,
> > > a strict calculation of dynamics of hexapeptide in solvent during 10
> > > microseconds took several months using a bunch of powerful computers.
> > > In real life proteins consist of roughly 100X more aminoacids and fold
> > > over at least 10,000X longer periods. Power requirements grow
> > > exponentially with increasing number of elements.
> > >
> > > DK
> > >
> > > >Artem Evdokimov wrote:
> > > >>
> > > >> Well it's easier to ask what do we know in this case, than what we do
> not.
> > > >> Even if we think of solving the folding problem by sheer brute
> force - i.e.
> > > >> by computing every interaction between the polypeptide chain,
> solvent, ions,
> > > >> etc. we still do not exactly know how to describe individual
> interactions.
> > > >> The best tool to do that is, undoubtedly, quantum mechanics because
> > > >> 'molecular dynamics' is, essentially, utter disaster - at least until
> some
> > > >> clever postdoc discovers a way to describe the interatomic
> interactions that
> > > >> does not involve rubber bands and bouncy balls on strings.
> > > >> Even if there were computers big enough to calculate all the
> necessary
> > > >> parameters we are still left with a question of how deep does our QM
> > > >> calculation has to be in order to succeed. It has been demonstrated
> many
> > > >> times that if one does not use an orbital model that is complete
> enough or
> > > >> if one does not use total correlation, etc. then the results of QM
> > > >> calculations are dubious. It has been shown that in many cases DFT
> (density
> > > >> functional theory) can save one a lot of processing time but even DFT
> > > >> calculations for something as big as even a modest peptide are, at
> present,
> > > >> impossible. If you add solvent atoms and the need to compute things
> until
> > > >> they converge (which, in case of folding can be very long in
> computational
> > > >> terms) then the problem becomes truly scary.
> > > >> Then, there's a question of how does the folding of the polypeptide
> depend
> > > >> on the ribosome from which the nascent peptide chain emerges, and on
> the
> > > >> chaperones and other cellular components. Yes, there are proteins
> which are
> > > >> known to fold by themselves, but many (I am inclined to guess that
> most)
> > > >> proteins aren't that easy.
> > > >>
> > > >> I am sure that if you ask a number of scientists in the field there
> will be
> > > >> conflicting opinions, deviating in both the optimistic and the
> pessimistic
> > > >> direction from mine. However, I am equally sure that a number of
> people will
> > > >> express similar if not identical (rather pessimistic) outlook on the
> current
> > > >> situation in folding simulaitons.
> > > >>
> > > >> Cheers,
> > > >>
> > > >> A.G.E.
> > > >>
> > > >> "Tim Davies" <daviest at shaw.ca> wrote in message
> > > >> news:3C307889.1C6CDF9 at shaw.ca...> > > >> > thanks for the information. i had no idea that the math was not
> > > >> > understood.
> > > >> > What information could be missing?
> > > >> > thanks
> > > >> > tim
> > > >> >
> > > >> > Artem Evdokimov wrote:
> > > >> > >
> > > >> > > No. The problem isn't just mathematical - it is fundamental - the
> > > >> science
> > > >> > > does not have all the relevant information, so the folding models
> that
> > > >> we
> > > >> > > build are incomplete.
> > > >> > >
> > > >> > > A.G.E.
> > > >> > >
> > > >> > > "Tim Davies" <daviest at shaw.ca> wrote in message
> > > >> > > news:3C2F97C9.F33EEC26 at shaw.ca...> > > >> > > > I am wondering if there are any known sets of equations which
> > > >> completely
> > > >> > > > predict folding and resulting structure. The calculation
> complexity is
> > > >> > > > not relevant just the interest to learn if there is in fact a
> set of
> > > >> > > > math which describes the phenomenon accurately
> > > >> > > > Thanks
> > > >> > > > Tim Davies
