From: Eugene Leitl (eugene.leitl@lrz.uni-muenchen.de)
Date: Tue Dec 14 1999 - 16:11:08 MST
hal@finney.org writes:
> It would seem that we have a ways to go before attempting this. We not
> only need the 3D shapes, we need to know the chemical affinities of
> each portion, and the mechanical properties. As a trivial first step,
> we'd need to know exactly how ribosomes perform protein synthesis.
> Very little is known about this at the mechanochemical level. They only
> obtained a detailed 3D map of the ribosome within the past few months.
> It appears that it may be something of a clockwork mechanism, with
> internal moving parts. I can't wait until we can really understand this
> molecular machine in detail.
We seem to have a confusion of sorts here, about what it means to
simulate a cell to be able to understand it.
In the field of molecular modelling for instance, there is a different
degree of simulation detail, known as the level of theory. Typically,
at a deeper level of theory, you tend achieve a higher accuracy at a
much higher associated computational cost. For instance, because of
how current QM codes scale, it is computationally infeasible to model
anything but the active site of the enzyme at that low level of
theory. So current most advanced models are hybrid, using classical
molecular dynamics/mechanics for the rest of the enzyme, feeding the
geometry into the quantum treatment of the active enzyme core.
Frequently, it is possible to abstract a given system, by stepping
back, tilting your head, squinting a bit, and discarding its aspects
we currently don't need. An enzyme can be thus seen as a molecular
contraption, processing freely diffusing species A in a given cell
compartment into other species B. Of course it is more complicated
than a mere Michaelis-Menten kinetics plot: enzyme's activity can be
modulated by other means, it may be immobilized at a surface, the
educts/products will be most likely be actively transported, but you
got the general idea. No need at all to go back to the molecular
dynamics level, or even (godforbid) formidable Herr Schroedinger.
Notice that it is sometimes might be necessary to look down to a
deeper level of theory to calibrate a higher level model. For
instance, one might not only use experimental data to obtain
parameters for a MD system, but try to extract them from
QM. Currently, afaik this cannot be done purely
automatically. However, I am confident that such automatically
learning forcefield codes are feasible in principle. In fact solving
the Protein Folding Problem (PFP) will probably require such codes,
since, as I already mentioned, our current forcefields are not
accurate enough to meet the challenge.
Similiarly, we can consider the ribosome as a black box which takes
different inputs to produce a given output: an amino acid string.
In a sense, most vanilla cells also don't do anything magical, they
might thus be modellable as a mechanical system with a rich inner
state, capable of sending and reacting to signals.
Of course there is a special class of cells of dear interest to us
all:
http://www.google.com/search?q=neuron+simulation&num=100
Today the molecules, tomorrow cell assemblies?
> As we learn more about these protein machines I think this will add
> further interest in the prospects for nanotech. If proteins are just
> enzymes, with specially-shaped active sites, that is old-fashioned
> chemistry. But if they are mechanical devices, Drexler's gears and
> cables don't look so outlandish.
True. In fact I had my mechanosynthetic epiphany when I tried to
understand as to why life does so much better than the organic
synthesis in the lab. It is about control, and machine-phase chemistry
is about as much control as it gets (Rejoiceth, you control freaks!
For your Kingdom is at hand ;)
This archive was generated by hypermail 2.1.5 : Fri Nov 01 2002 - 15:06:04 MST