[BioC] Newbie needs help with exprset

[Seth Falcon]

Hi Sarah,

You may find the ExpressionSet tutorial presented at last October's
BioC intro course to be useful:

  http://bioconductor.org/workshops/2006/biocintro_oct/labs/ExpressionSet/

Have a read through ExpressionSetIntro-solved.pdf.  It walks through
the details of building an ExpressionSet object from scratch.

Naomi Altman <naomi at stat.psu.edu> writes:
> What we do is to to use read.table or read.delim 
> to read the data into a dataframe.
>
> I have not found it necessary to create an 
> exprSet object from the dataframe.  

Actually, I think there are a lot of advantages to getting your data
into an ExpressionSet (that's the new and improved version of
exprSet).  Here are a few reasons:

  * Automatic/enforced alignment of pheno data (e.g. clinical vars)
    and expression data.  If your pheno data gets misaligned, your
    analysis will be compromised.

  * Easy subsetting of genes (probe sets) and/or samples.  Again, once
    you've aligned the data, you don't have to think about it again.
    If you have separate data.frame or matrix objects for the
    expression values and your phenotype description data, you have to
    do this manually for each subset you want to explore.

  * Many Bioconductor packages make it easy to apply analysis
    techniques if what you have in hand is an ExpressionSet.  For some
    examples, browse recent BioC workshops on the website.

> However, if you feel it is necessary, you can create each component
> using the columns of the dataframe as needed and then set the class
> to exprSet.

This is a bad idea and is not guaranteed to work.  Your next
suggestion is much better...

> The documentation for an exprSet is in Biobase.  (Note the uppercase
> S.)  It shows how the "new" command can be used to create the
> exprSet once you have all the components.  (Components can be
> empty.)

However, exprSet is being replaced by ExpressionSet, so it would be
better to start there.

Best,

+ seth