Expression analysis

[A.Domingues]

Hi,

I am doing analysis of a few MJAY arrays and want to understand the
relationship between splicing and expression. For this I need to
generate files with expression changes (and corresponding statistics)
for all the genes present in array regardless of whether they are
significantly changed in my treatment or not.

The question is: how to do it? I've tried to change the DABG p-value
and "remove probesets" thresholds but how low can I go to include as
much as I can? Also, can this type of analysis be done also for the AS
events?

Removing very low (or absence of) expression would be acceptable.

Cheers.

[Nathan Salomonis]

Greetings,

It sounds like you would like to view alternative exon statistics for all
genes as opposed to just those reported as "significant" by AltAnalyze. For
the gene expression analysis, this is done by default, but for Alternative
exon analyses, only exons and corresponding genes that were called as
alternatively expressed are reported in these standard summary files:
http://code.google.com/p/altanalyze/wiki/AnalyzingASResults

One important point to clarify, is that all alternative exon analyses are
performed at the exon- and reciprocal junction-level. Hence, any summary of
these statistics at the gene-level must examine all exon/junction statistics
for that gene, that are reported as changed.

If you select the option "Export all normalized intensities", under
"Advanced Options" of the "Alternative Exon Analysis Parameters" window, you
will get a text file saved to AltResults/RawSpliceData with all normalized
expression values for all array comparisons (exon and reciprocal-junction).
This will not be gene-level, but these identifiers can be manually
summarized to this level. Alternatively, you can attempt to results for all
alternative exons and reciprocal junctions in the main summary files,
including the gene summary file. Directions for this are given below when
using AltAnalyze in the command-line mode. These options can be mimicked in
the user-interface as well:
http://code.google.com/p/altanalyze/wiki/ReturnAll

These results will still not contain alternative exon statistics for all
genes, as some genes will have only one probeset/gene and thus are not
amenable to alternative exon analyses. Please tell me if you need additional
clarification.

Best,
Nathan

[Nathan Salomonis]

As a follow-up note, when-ever you change the expression threshold
parameters (dabg and expression cutoff), you effect the exon and gene-level
results by incorporating more or less probesets in to the analysis which
effects the gene expression in the two conditions analyzed and adds
additional non-expressed probesets that can be called as significant.

For example, if you set the DABG p-value threshold to 1, all constitutive
probesets will be used for averaging to obtain a gene expression value, even
those that are likely not expressed. This can effect your normalization
signal and result in differences between the expression pattern of your
exons and genes (or loss of these changes). If a non-expressed probeset is
introduced into the analysis, when gene expression changes are present,
these probesets will be reported as alternatively expressed, since their
pattern will differ from the constitutive probesets (gene-level).

Hence, we recommend incorporating the default filters and reporting any
genes not called as alternative as "NA", for most user analyses. Again,
choosing to export all normalized intensities will still give you all other
analyzed gene statistics as well, but will still comprise a subset of all
genes on the array (since many will not be deemed expressed).

Best,
Nathan

On 8/29/11 9:25 AM, "Nathan Salomonis" <nsalo...@gladstone.ucsf.edu>
wrote: