Expression Correlation

[susana.b...@nasa.gov]

Dear all,

I am trying to use Expression Correlation for creating a network of
genes that exhibit of correlation in their regulation. Well, at least
this is what I thought that Expression Correlation would do for me
(please, correct me). Now I realize that I am not understanding the
tool.

I import my expression matrix (no problems this time) and run the
plugin. Could you give me a more clear explanation on how to use the
cutoffs? I read the manual from the Bader Lab, but I find it not very
explicit. I even played with their sample data so that I could figure
out some rules, but I was not successful.

In summary: my main three questions are:
a) Will Expr Corr help me to design a network/pathway from a
expression level file?
b) Does Expr Corr work even if I have just two conditions compared?
(therefore my expression data is just given in a single column
representing "fold change")
c) How do I work with the cut-offs?

Thank you very much in advance,

Susana

[Andrés Flórez]

Hi Susana,

I came up with the same questions some weeks ago and reading some
literature I got the follow conclusions (If somebody can extend this
answer I will appreciate it)

On Fri, Feb 13, 2009 at 3:47 PM, <susana.b...@nasa.gov> wrote:
>
> Dear all,
>
> I am trying to use Expression Correlation for creating a network of
> genes that exhibit of correlation in their regulation. Well, at least
> this is what I thought that Expression Correlation would do for me
> (please, correct me). Now I realize that I am not understanding the
> tool.
>
> I import my expression matrix (no problems this time) and run the
> plugin. Could you give me a more clear explanation on how to use the
> cutoffs? I read the manual from the Bader Lab, but I find it not very
> explicit. I even played with their sample data so that I could figure
> out some rules, but I was not successful.
>
> In summary: my main three questions are:
> a) Will Expr Corr help me to design a network/pathway from a
> expression level file?

Expression Correlation methods build a similarity matrix for nodes,
using a distance measure to decide when two nodes should be
associated and then link them by an edge.(Allan's comment)

> b) Does Expr Corr work even if I have just two conditions compared?
> (therefore my expression data is just given in a single column
> representing "fold change")

Some authors have recommended that it is necessary at least 8 samples
to analyze expression data looking for patterns.
I suggest you to read those:

http://bioinformatics.oxfordjournals.org/cgi/content/abstract/18/9/1184
http://www.biomedcentral.com/1471-2105/7/S4/S13

Hope this helps,

> c) How do I work with the cut-offs?
>
> Thank you very much in advance,
>
> Susana
>
>
> >
>

--
Andrés F. Flórez

Associate Researcher
Program for the Study and Control
of Tropical Diseases (PECET).
Universidad de Antioquia.
Medellín-Colombia.
Carrera 53 # 61 - 30, SIU, Lab 632
www.pecet-colombia.org
Phones: +57-4-219 6502 / 219 6622
Fax: +57-4-219 6511
Cell phone: 3103462268
e-mail: andres...@pecet-colombia.org

[swaraj basu]

Hello,
         yes expression correlation will help you in making a pathway, but first you have to load names of your genes or IDs for which expression is mapped in form of an excel sheet into Cytoscape. Then you load your expressionvalues as a text file with same IDs and corresponding expression in rows, separately. Go to expression correlation plugin --> advanced --> gene network: Preview Histogram, and the plugin builds an initial histogram. If you are successful uptill this level please do inform me and I shall help you further with the cut offs.
                                                        Swaraj

[ophth1]

Hi!
we have similar problems, too. We would like to use the expression
correlation plug-in in order to create cytoscape networks based on our
expression data and conditions. The authors of the plug-in state in
their read-me that it is possible to distinguish and compare
expression correlation networks furthermore based on conditions (e.g.
groups or disease-states etc.). But in their sample data file, their
are no conditions present. Does anyone know how to define these
conditions, how to construct the data matrix and how to handle them?
Furthermore they state that it might be easy to compare different
expression correlation networks using the DIFF cytoscape plug-in.
Where can I find this plug-in?
Thanks!

> > Susana- Zitierten Text ausblenden -
>
> - Zitierten Text anzeigen -

[ophth1]

hi!
I have a similar question: the authors of the plug-in state in their
help file that it might be possible to use conditions to compare
expression correlation networks amongst different groups or disease
states (conditions). However, in their example expression file their
are no conditions defined and their do not state how to define.
Does anyone know how to setup experimental data WITH conditions in
order to use them with this plug-in?
thanks!

On 22 Feb., 09:10, swaraj basu <projectb...@gmail.com> wrote:

[swaraj basu]

Hello,
          set up your expression data as follows
                 
                         
row1-gene1   column1-condition1     column2-condition2    column3-condition3
row2-gene2   column1-condition1     column2-condition2    column3-condition3
row3-gene3   column1-condition1     column2-condition2    column3-condition3


live ex

NC101  1.30   -1.45     1.67
NC102   0.00   -1.70    2.67
NC103   5.00  -1.78   3.40


all values should be tab separated. As far as different network comparison to be done, even without diff plugin using cytoscape Merge network facility it can be performed.
                                                      Swaraj

[Susana]

Thank you very much for your replies. Yes, I am able to load the
expression matrix and generate an histogram. I would appreciate your
help with the cutoffs. What do I do next?

Also, as Andres indicated in his reply, I should have several samples.
Would it work with just two samples which correspond to two treatments
compared to control? (so, my samples would be Control, Treatment 1,
Treatment 2, but as I am expressing my results as expression ratios
over control, there are actually only two columns with expression
values).

I appreciate your help.

Susana

> > Susana- Hide quoted text -
>
> - Show quoted text -

[piet molenaar]

And in addition to this answer: as Swaraj already states I think the authors mean the network merge plugin in stead of the diff plugin; this enables you to subtract two networks: tick the Operation dropdown and choose difference...
Piet

--
Piet Molenaar
p.mol...@amc.uva.nl
Department of Human Genetics, M1-131
Academic Medical Center
University of Amsterdam
Meibergdreef 9
1105 AZ Amsterdam
the Netherlands

tel (+31) 20-5666592
fax (+31) 20-6918626

[piet molenaar]

Hi Susana,
Probably it's wisest to contact the authors of the tool directly; these questions consider the specifics of statistical analysis of biological data and are therefore slightly out of the scope of this helpdesk that mainly focuses on the workings of Cytoscape as a tool.
Good luck!
Cheers,
Piet

[swaraj basu]

Hello Susana,

                     sorry for the delay, once you get the histogram, you must choose a cutoff on both sides that is positive and negative, ideally the default cutoff is fine that is 0.95 but if u want to be more stringent then you can use 0.99 or 0.90 if you want to be less stringent. I would suggest from my personal experience that since you have 2 columns you should be stringent and use 0.99 as cutoff. I also agree with Piet that you can further refer to the biological significance by directly contacting the authors or you can refer any paper related to expression correlation either pearson or spearman and it will help you.

                                                                                                                                    swaraj

[lani.lic...@gmail.com]

Hi,

I have a separate question regarding Cytoscape 2.6.0 and Expression
correlation plugin.

I have loaded a file of IDs, and expression matrix successfully (the
exp values show up in the node attribute browser).

However, when I do Plugins-> expression correlation network ->
construct network (or any other selection from this sub-sub-menu),
nothing happens.

The plugin manager lists expression correlation plugin as installed,
and it appears in the menu bar.

This could either be a problem with

(1) connecting to a web server (is this even required for expression
correlation?)

(2) a plugin in problem. My "Manage Plugins" window footer states
"Error: failed to read xml file". But, the plugin IS installed...

(3) a problem with my network file, but I tried the plugin for a
variety of networks (including sample data) and it does nothing.

Thanks!

On Mar 12, 12:34 am, swaraj basu <projectb...@gmail.com> wrote:
> Hello Susana,
> sorry for the delay, once you get the histogram, you
> must choose a cutoff on both sides that is positive and negative, ideally
> the default cutoff is fine that is 0.95 but if u want to be more stringent
> then you can use 0.99 or 0.90 if you want to be less stringent. I would
> suggest from my personal experience that since you have 2 columns you should
> be stringent and use 0.99 as cutoff. I also agree with Piet that you can
> further refer to the biological significance by directly contacting the
> authors or you can refer any paper related to expression correlation either
> pearson or spearman and it will help you.
>
> swaraj
>

[swaraj basu]

Hello,
         Most probably it is an error in ur plugin installation, remove the plugin and install again...I hope it will work

[Ilana Lichtenstein]

thanks for the tip. However I have tried the installation on a second computer and it also  "does nothing" when I select the plugin plugins->Expression Correlation Network " -> Construct Correlation Network" from the menubar.

 

I am extracting the jar file to the cytoscape/plugins directory.

Do I need to 'activate' the Jar file in some way? It is unusual as my other plugins work.

 

 

Ilana

[swaraj basu]

Hello,

          I think once you have your network loaded and the expression values uploaded as a pvals files by pressing Ctrl+E then it should work fine. The names of the network nodes i.e their IDs and the names in first column of pvals file should be the same.

                                                                                     Swaraj