[BioC] Several questions goCluster and GOstats...

[Gunnar Wrobel]

Hi Lourdes,

> 1. How can I change configuration parameters from a goCluster object
> without running again config(goCluster_object)?  Alternatively, where
> could I find documentation about doing this? For example, I would like
> to try different similarity measures  or  different pvalue cut-offs,
> etc.
> In the manual, it is showed how to change the visualization method so
> I assumed there is a slot for all other parameters.
>

You can use the "setup" function. This function retrieves a
configuration from a goCluster object as a list. This configuration list
can be modified and then assigned to a new goCluster object.

> a <- setup(goCluster_object)
> a$sign$threshold <- 0.1
> test <- new("goCluster")
> execute(test) <- a

Instead of directly executing in the last step you can also assign the
configuration using "setup" again so that you can run the analysis at a
later time:

> setup(test) <- a

> 2. goCluster will show only the most significant GO terms (with the
> lowest p-value) if there are overlaps.  Is there a way to see all
> selected GO terms  (and their p-value), similar to GOHyperG output?

Yes, in the significance slot.

> goCluster_object at sign@selected

This list is splitted according to the different clusters that were
identified (you'll find those in goCluster_object at algo@clusterset)

> 3.  Suppose I use GOHyperG to select the GO terms for a  determined
> set of genes,  can I use goCluster for visualization? if yes, how?

Actually using GOHyperG should result in the same values as goCluster
since I used the GOHyperG code for verification. I'll add a section in
the users manual that will verify that.

But in principle nothing prevents you from replacing the statistical
module as detailed in the developers manual of goCluster. The package is
meant to be extensible. In case the developers manual is not clear
enough (guess it might not ;) ) don't hesitate to contact me.

> 4. Is the code for the examples described in the vignette "Using GO
> for statistical analyses" available?

It is part of the GOstats package so you should have a file in your
R-library:

R/library/GOstats/doc/GOstats.Rnw

Cheers

Gunnar