automatic planning for molecular cloning workflows?

[Patrick O'Neill]

Hello all,

There's an old joke that professors prefer to use an extremely
high-level language called gradstudent. Gradstudent allows you to
write notes in English that get translated into a week's worth of work
in the lab for the hapless "interpreter". Having "executed" some
gradstudent code myself, I wonder: how far off is this technology?
The topic of scientific robots is a popular one on this list, but I'm
more interested in the issue of workflow planning: can you take a
natural language command like "amplify insulin" and translate that
automatically into a detailed protocol?

It sounds like an impossible problem, but there appears to have been
some work on it fairly early on. Beginning several decades ago there
was a lot of research into planning as a topic in AI, and programs
like MOLGEN and INFERGENE used molecular biology as a problem domain.
The researcher would describe a goal and a set of constraints, and the
programs would crank out an experimental design that yielded the
desired product given the constraints.

What ever happened to this stuff? The first generation of programs
was hampered by the problem of large knowledge-bases, which restricted
them to trivial experiments. But dealing with large knowledge-bases
is a problem we've gotten a little better at in the last thirty years.
Is anyone familiar with more recent efforts at this problem? Does
anyone use anything that sounds vaguely like what I described?

thanks in advance,
Pat.

[Bryan Bishop]

On Thu, Mar 15, 2012 at 9:18 PM, Patrick O'Neill
<synapse.a...@gmail.com> wrote:
> The topic of scientific robots is a popular one on this list, but I'm
> more interested in the issue of workflow planning: can you take a
> natural language command like "amplify insulin" and translate that
> automatically into a detailed protocol?

you might be interested in reading some stuff from the archives:
http://groups.google.com/group/diybio/browse_thread/thread/19df187ef1014735
http://news.ycombinator.com/item?id=2191105
http://88proof.com/synthetic_biology/blog/archives/290
http://groups.google.com/group/openmanufacturing/browse_frm/thread/a8d8ee245aaae97d#
http://groups.google.com/group/openmanufacturing/browse_frm/thread/b2e21ccc953d6328#

The language technology is easy. What I would like is to then
automatically generate a microfluidic lab on a chip that implements
the protocol, or generate beautiful documentation for the workflow
that I could give to a team of people (or, er, just myself). But most
people just use gradstudents, yes...

> It sounds like an impossible problem, but there appears to have been

doesn't sound impossible

> What ever happened to this stuff?  The first generation of programs
> was hampered by the problem of large knowledge-bases, which restricted
> them to trivial experiments.  But dealing with large knowledge-bases
> is a problem we've gotten a little better at in the last thirty years.

problems with knowledge-based stuff:
1) figuring out a good format/represetation
2) collecting it without killing ourselves in the process
3) finding something useful to do with it (sometimes this can pay for
the other two steps)

>  Is anyone familiar with more recent efforts at this problem?  Does
> anyone use anything that sounds vaguely like what I described?

biocoder ain't it

- Bryan
http://heybryan.org/
1 512 203 0507
irc.freenode.net ##hplusroadmap

[Mackenzie Cowell]

A simple way to play with this idea would be to establish a small controlled vocabulary for describing the inputs and outputs to various protocols.  Meta-protocols could be created by looking for a path from a desired input to a desired output.

Input: 10 ug 1uM plasmid DNA (BBa_I20259)

output: 400 Relative Fluorescence Units, 500nm.

I think a modest collection of protocols would be useful from an educational standpoint by helping users quickly grasp the informal grammer of available operations.

If the vocabulary of available biological operations (verbs) and inputs/outputs (nouns) became more structured, it could form the basis of a biological "test-driven development" system.  First, desired biological outcomes would be expressed using the vocabulary.  Then the biologist would execute an automatically-generated protocol until the tests passed.

I know dozuki.com (powers ifixit.com and makeprojects.com) allows authors to describe the Bill of Materials and Tools needed to accomplish a guide in a structured way.  Example: guide to making moss graffiti: http://makeprojects.com/Project/Moss-Graffiti/2033/1.  However, I don't think they structure the output of guides, so it's not possible to search for paths through their set of guides.

Mac

--
You received this message because you are subscribed to the Google Groups "DIYbio" group.
To post to this group, send email to diy...@googlegroups.com.
To unsubscribe from this group, send email to diybio+un...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/diybio?hl=en.

--
+1.231.313.9062 / m...@diybio.org / @100ideas

[Mackenzie Cowell]

I think Dozuki expresses their guides in oManual XML.

http://www.dozuki.com/api/0.1/doc

It would be interesting to try expressing a couple protocols in oManual.  http://omanual.com/.

Mac