Page 67 - Visions of the Future Chemistry and Life Science

56 J. M. GOODMAN

However, it is the interaction between these field which gives the best
chance of computers becoming the world’s best synthetic chemists.
Chess is not solved, in the way the simple game noughts and crosses
is solved, because the outcome of every game is not completely predict-
able. However, computers will usually win. In the same way, it may not be
necessary for computers to analyse all possible routes to a molecule to be
best at organic synthesis. It may be enough simply to be successful at
finding good routes. This makes the problem much easier, if it is assumed
that there are many good routes. The computer would begin by guessing a
route, and if it did not work, partially retracing steps, and trying again, thus
reusing the information which had already been gathered or calculated so
far as possible. Thoroughly exploiting the information that was developed
with each potential synthesis would be a crucial step. The time required
for conformation searching is dramatically reduced, if similar molecules
have already been investigated. For example, PM-toxin has a very compli-
cated potential energy surface, which may be searched directly by tradi-
tional methods, or which may be mutated from the conformation search
of an alkane, which is easier as it is particularly susceptible to a genetic
algorithm based approach. A web page illustrating these results is avail-
able: http://www.ch.cam.ac.uk/MMRG/alkanes/ Some more technical
details and a more complete list of references can be found in the
Millenium issue of Philosophical Transactions of the Royal Society
(Goodman 2000).
Will this allow syntheses to be automated? It depends how difficult
syntheses are (and will provide a way of quantifying this). It may be that
the best possible synthesis is not required, provided that a good route is
available, as assessed by total cost (including waste disposal and safety pre-
cautions), by time required, by certainty of success, by ease of using robots
to follow the procedure, and so on.
Organic synthesis is, and will remain, a very demanding discipline.
Brute force methods of calculating new synthetic routes will not be fea-
sible for a very long time, and pure literature based methods will also be
very time consuming, and will be restricted by the data available. An
hybrid approach provides the best hope for designing a synthetic machine,
and it is likely that such programs will become increasingly useful in the
new millennium. Most of the elements of these programs are available
now, but they are not sufficiently useful that they are an essential part of
every chemist’s work. An exhaustive solution may not be possible, so it is