ADL: autodock 3.05 on linux cluster?

Daniel Jana dfjana at gmail.com
Wed Nov 21 03:05:24 PST 2007


On 21/11/2007, Kalidas Y <kalidas.y at gmail.com> wrote:
> Please refer to the parallel autodock:

Just out of personal interest... do you use this? Do you see any
benefit? I do not use Autodock myself but I am in close contact with
users that do use and their autodock jobs.

In most of their calculations an Autodock calculation takes a few
hours to finish. On the other hand, for ab initio calculations I am
used to wait days to weeks (one of our users had to wait three months)
for a result! In a situation like this, it seems reasonable to use
parallelization. If you need this result to proceed your calculations
you cannot wait a six months (or one year) for the job to finish. But,
for Autodock calculations I can't see much gain on doing such a
parallelization.
There is one case in which it may be useful, if what you want to
screen a library of compounds. But, for that Autodock also has way of
proceeding by doing a more shallow test, which separates the good from
the bad instead of (more) accurately saying which is/are the best.

So, the place where I want to get to is. Do (any of) you see the
*real* need for a parallel Autodock? Could you elaborate a bit, just
for me to understand...

Thank you very much in advance,
Daniel

PS - Keep in mind that software, to run on a cluster, does not need to
run on multiple processors for each job. You can put it working for
each job with one processor and there is no limit on how many
instances of Autodock could be running in any one cluster.
PPS - If anyone wants to answer privately I'm ok with that, especially
if the person is willing to reveal a little bit more, privately, then
to the mailing list. However I think that this could be an interesting
subject to be discussed publicly, to see the real interest on a
parallel implementation of Autodock. After all, it comes from Scripps
in a serial version only...


More information about the autodock mailing list