[BioC] GAGE vs other GO analysis tools

[January Weiner]

Dear all,

I am trying to use GAGE for GO and pathway analysis, but the results
of the GAGE analysis applied to data sets that I know rather well are
strange.

The samples come from patients suffering of an infectious disease, and
if compared with controls, they normally show an enrichment in GO
terms and KEGG pathways related to immune answer.

In GAGE, the same data also show a significant enrichment, but to
ribosomal functions, for example:
              p.geomean stat.mean        p.val        q.val set.size
                                       name
GO:0005840 1.610041e-10  5.917761 1.614836e-58 4.925249e-55      185
                        GO:0005840 ribosome
GO:0003735 3.190154e-10  5.893759 9.135429e-57 1.393153e-53      141
GO:0003735 structural constituent of ribosome
GO:0006412 1.051080e-09  5.581083 1.619059e-54 1.646043e-51      351
                     GO:0006412 translation
GO:0030529 1.412392e-09  5.339373 3.149454e-50 2.401458e-47      408
       GO:0030529 ribonucleoprotein complex
GO:0033279 3.177217e-08  5.093234 7.376959e-43 4.499945e-40      109
               GO:0033279 ribosomal subunit
GO:0006414 2.438856e-07  4.695261 1.070581e-36 5.442121e-34       98
        GO:0006414 translational elongation

I can't believe the above; not only these results are not confirmed by
any other analysis (topGO, GOrilla, online GO analysis tools, GSEA,
SPIA for comparison with kegg.gs), but furthermore if one is to plot
the microarray intensities of the genes by group and by GO term for
the above GO terms, it becomes apparent that there is little
difference in the analysed genes.

I know that I am giving but few details in my e-mail, but I hope that
maybe some other person had similar troubles with GAGE.

Kind regards,
j.

-- 
-------- Dr. January Weiner 3 --------------------------------------