More regression models / Sigmoid regression / EC50

[GoogleCodeExporter]

Please describe the purpose and the intended behaviour of the new feature.
Most biological reaction follow a Sigmoidal/LogSigmoidal regression. 
Creating tool to calculate such regression such as EC50, Min, Max etc. 
would be great.

If it has connection to other parts please state it too.
I wanted to devellop such library but I don't know enough about KNIME to 
do it (in addition, I was having problem to get some of the HITS source 
code to use as a base).

For this I wanted to use Michael Thomas Flanagan's Java Scientific Library 
(http://www.ee.ucl.ac.uk/~mflanaga/java/ need to get permission to 
redistribute) which provides the tools to calculate Ec50 etc. An other 
resource that could work is: www.ncgc.nih.gov/pub/openhts/curvefit.

Note that it would be nice to be able to get this EC50 on sets with 
replicates and for multiple features measured.

For instance if you have a table as follow:

CompoundID;[Concentration];Value1;ValueX;etc
A;1;3;5;...
A;1;2;4;...
A;10;15;30;...
A;10;18;27;...
B;...

to obtain an EC50 (or other parameters of the sigmoid) for each of the 
compounds.

I think part of this can be done using KNIME nodes provided with Hits, but 
the Sigmoid or other type of regression and there plotting does not exist. 
Perhaps this could be done?

Original issue reported on code.google.com by gbon...@gmail.com on 22 Apr 2009 at 12:32

[GoogleCodeExporter]

Thanks for your interest. I guess this would take about 2 weeks to create such 
nodes
using that library. But it would take much more time to document, create use 
cases,
help and other supporting materials. So currently we do not plan to add these
features (at least not in the near future), our focus is on siRNA screens.
If there will be increasing need to this kind of functionality we will create 
these
nodes.

You had problems getting the source code of the projects, or you had problems 
with
modifying to your needs. The latter seems more probable, the computations are 
mostly
done with cellHTS2, which is written in R (so the framework in Java help not too
much). For the former you might want to take a look at
http://code.google.com/p/hits/source/checkout .

You might ask similar question on KNIME forums, it has a good chance someone is
already working on this.

Original comment by aborgabor on 8 May 2009 at 3:26

• Added labels: Component-Logic

[GoogleCodeExporter]

Thanks for the answer. I will ask at the KNIME user forum.

Original comment by gbon...@gmail.com on 8 May 2009 at 9:44