[ome-devel] data grouping/searching, creating dataset

[Bernd Jagla]

Harry,

 

Thanks for your interest in our work. 

Bernd:

 

Interesting stuff: some comments below.

 

On Nov 23, 2005, at 2:21 PM, Bernd Jagla wrote:

Then we want to display a particular feature for all the wells/plates as a
heat map to identify by eye any biases or problems in the data set. For this
we envision right now using Partek's Screener Solution which could interface
with OME through ODBC. A simple (complex) select statement should bring all
the necessary information into the Screener Solution.

 

As a data visualization guy, I'm interested in what you're doing here - can
you provide more detail?

When you say heat map, I assume that the feature for the wells/plates would
be something that has a numeric value? Do you want to do dendrograms, or is
outlier detection your main task?

 

At the moment I am just looking for plotting the numerical values we came up
with from analyzing our images. So the focus for me is data integration and
getting things going. We don't even have any valuable data yet. I will keep
you informed of where we are going.





Then we have to select hits and analyze the screen: 

Display a particular feature for all wells and plates. (This is for quality
control). 

Select a relevant feature and a threshold.

Write back the results (somehow) to OME. Those results will eventually group
the results in different categories (hit, non-hit, not-known, positive
control, negative control, etc). This actually still has to be solved.

Custom STs, or CategoryGroup/Category annotations can help. 

 

Do you mean we should generate an ST for each group, or bypass the given
dataset structure? I will have to look into the CategoryGroup/Category
annotations a bit more.





Then we want to see/ cross-link the results with other screens: of these
hits, which other assays where they active in? Which compounds where active
in a given set of screens? Give me a complete profile of all the compounds
and in which assays they where active, given a specified threshold etc.

 

when you say hit, do you mean image? ie, which other datasets/experiments is
a given image in? Also, can you clarify threshold?

 

A hit is defined by a value derived from an image being higher than a
predefined threshold. The threshold is usually set such that we get a
certain number of hits.

Images are confocal microscope images of cell-based assays that were treated
with different compounds.





Of course at some point we want to see the compounds with the results and
everything but this will wait a while and needs probably further discussions
on how to integrate all this.

right.



Then we have to push some of the data to PubChem at the NIH. How could we do
this? A simple script with select statements could do the work, no? Maybe we
can write a chain module that automatically propagates the data to PubChem
(they use XML - their own format of course).





 

One possibility would be to use OME's exporting facilities to generate OME
XML files and then XSLT to translate into PubChem's schema.


I am actually having problems with exporting my images and information, but
I have to admit that I didn't do much about it yet. I just tried the export
image function which didn't work.

 

At the moment I believe that the most important part is to improve (or for
me to learn how to do) the grouping features of OME. I don't think it is
good that you have to know in advance that you want to add something to a
data set. From my experience I would rather play with the data and mark e.g.
bad wells along the way, select images that look interesting in one way or
the other and say: Wow they look cool lets make a new dataset and collect
them here, or lets add them to the list of interesting data. As far as I
know this is only possible once you know the name of the data set, and then
you have to search for them again.

there are lots of ways to think about this sort of functionality - we've
discussed pieces of it. The closest thing that is implemented is the
category group/category annotation tool, but that does require starting from
a selected set of images.

 

Also for the sake of better understanding can you please point me to all the
relevant information concerning data sets: how to programmatically do it,
how to do it with ODBC, and the web interface?

The best places to start from are in the "Developer's documentation" section
of the web site links.  We can give more feedback for more specific
questions.

 

Ok, I will try to start there (after I have read in the results correctly).

 

 

Thanks a lot,

 

Bernd

 

I would be more than happy to explain in more detail or discuss our
strategies as we are just in the process of getting started.

 

Thanks a lot for all your kind help.





You're very welcome...

-harry

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.openmicroscopy.org.uk/pipermail/ome-devel/attachments/20051129/74dec228/attachment-0001.htm