Re: processing power and chemistry

From: Anders Sandberg (asa@nada.kth.se)
Date: Sun Jun 15 2003 - 17:09:46 MDT

  • Next message: Robert J. Bradbury: "Re: processing power and chemistry"

    On Sun, Jun 15, 2003 at 05:09:13PM -0500, Kevin Freels wrote:
    > The basic properties of all of the elements of the periodic table have
    > been know for some time. If we had a computer with sufficient
    > processing power, couldn't we have simulation software that could tell
    > us pretty much every chemical compound/material that could be made?
    > This would eleminate the trial and error approach and solve possible
    > every disease!

    The problem is combinatorics. The number of chemicals that you can make
    increase very fast with the number of atoms. For ten carbon atoms there
    are 75 different alkanes, for 20 366,319, for 30 4,111,846,763 and for
    40 62,491,178,805,831 (from http://www.rod.beavon.clara.net/quiz.htm).
    And that was just combining carbon and hydrogen. The real problem is
    that each bond in these molecules can also turn in different directions
    so you end up with a number of configurations growing roughly
    exponentially with size for *each* possible molecule. So finding which
    chemical in which low-energy configuration would fit a receptor or have
    a neat property would require searching through a space that increases
    superexponentially with the number of atoms involved. This is
    impractical given most advances in computing, and even quantum computers
    would only make the problem very expensive.

    > I began thinking about this when I was watching a program that
    > discussed how snake venom was being used to develop treatment for
    > diseases. I estimated that there are so many possible chemicals out
    > there in nature that we could probably never get to a point where we
    > could try them all on everything.

    Yes, isn't it annoying?

    > I know that we currently don;t have the processing power to do this,
    > but does anyone on this board know roughly how much processing power
    > we would need?

    Let's say that the number of chemicals containing N atoms selected from
    a repertoire of 92 are roughly (92^N)*(N!) (select the atoms, order them
    in some order; this is an overestimate). The atoms have on the order of
    N bonds which can have ~7 degrees of freedom (length, direction,
    torsion), so for each molecule we have to search through a 7N
    dimensional space to look for minimum energy/fitness maximum
    configurations. So the total amount of work of finding the best by brute
    force is on the order of (K^N)(N!) for a large K - very slow. Even an
    exponential speedup only turns it into a N^2 log N search with a bad
    constant.

    -- 
    -----------------------------------------------------------------------
    Anders Sandberg                                      Towards Ascension!
    asa@nada.kth.se                            http://www.nada.kth.se/~asa/
    GCS/M/S/O d++ -p+ c++++ !l u+ e++ m++ s+/+ n--- h+/* f+ g+ w++ t+ r+ !y
    


    This archive was generated by hypermail 2.1.5 : Sun Jun 15 2003 - 17:15:35 MDT