From: Eugene Leitl (eugene.leitl@lrz.uni-muenchen.de)
Date: Sat Jan 29 2000 - 20:57:20 MST
Technotranscendence writes:
> > We first need a good water model. I haven't modelled freezing in large
> > volumes (it's on my todo list), but purportedly the freezing point is
> > totally off-key.
>
> Please explain. (If it will take too much space here, please email me
> directly.)
The freezing point in machina is far from ~273 K. Of course it could be
microdroplet supercooling, but one find that out by seeding the volume
with an ice nucleus. As I said, I haven't tried it out yet.
> > Virtual screening is very new, and requires a lot of resources, both
> > in terms of hardware and manpower.
>
> But genetic algorithmns have been around since, at least, the early 1970s.
> If one could start with a simple model -- one that hardly works -- and just
> build on it, I bet this could be done a lot quicker. We don't need
Structure generation is more or less trivial. You can generate
hundreds to thousands of structures per second. Screening them
virtually is much more difficult, especially if things go beyond
"mere" docking. Freezing in presence of cryoprotectantcs is probably
synergistic (the cryoprotectant influences the ice structure and the
growth mechanisms). If you assume it isn't, you might produce
garbage. Off the top of my head, a virtual screening run with
simulated freezing with current hardware (64 node Beowulf cluster) can
take around a week for a single candidate. Maybe more than a week.
> perfection here. After all, this would only act as an input into actual
> physical testing. The searches coudl act to weed out more bad
"only"? The computer suggests a structure. You can't buy it, so you
make a literature search (if you're lucky, you can find a procedure,
order the chemicals, and hopefully synthesize the thing in a few
weeks). If not, you have to develop a synthesis route first. Then you
have to purify it, and test it. (Testing can be done in a half a day).
> > Cryobiology is but a tiny subfied of a subfield, counting about 200
> > practitioners world-wide, most of them past their prime. Molecular
> > modelling is hence slow to enter the field.
>
> I can understand, but that also means no one else is probably doing the work
> there, so you need not fear competitors for a while. You'll be the pioneer.
You still need the resources, and the manpower.
> Also, cryobiology is not the only field I mentioned. Antiaging research is
> the other. What of that? If one can develop a good model of aging at the
> cellular or molecular level, then finding new ways to combat it would seem
> to me to be a lot easier. From the looks of it, it appears most research in
> this area is identify a mechanism, find something that inhibits it, look for
> similar things to that inhibitor, and so forth. Am I right?
>
> A simulated aging cell could be used to more efficiently search for aging
> inhibitors and therapies, no?
We cannot even fold proteins yet. Virtual cell at molecular resolution
is future technology.
> Specifically, I'd like to know how people come up with ideas for new
> cryoprotectants.
Educated guess, blind trial and error, and rational design. A little
bit of everything. You can also go for combinatorial chemistry and
evolution in vitro.
This archive was generated by hypermail 2.1.5 : Fri Nov 01 2002 - 15:26:32 MST