I'm working on optimizing DOCK6 on a massively parallel computer, and
we're running into a problem with scaling. We'd like to find a way to
estimate the time it takes to dock various ligands in a query. This
information can be used to produce an ordering that allows all the
compute nodes work efficiently without a lot of idle time.
We have a start already: the larger the molecule, the longer the
calculation, in general. This rule only results in limited success.
There is some other feature of the molecule or computation that causes
some small molecules to take an exorbitant amount of time, and some
larger molecules are finished very quickly.
Does anyone know of any similar work being done? Any suggestions or
comments are welcome. (Also, is there another usenet group that might
be a better venue to ask in?)