Dear Community,

based on some initial in vitro experiments, and a subsequent cancer microarray dataset analysis in R, i would like to perform some gene-set tests, for specific pathways and ontologies, regarding my phenotype of interest. Briefly, based on a two-group condition, we are mostly interested in identifying biological processes related to neutrophils, and subsequently more generally to inflammation. So the two major approaches under consideration:

A) Have identified through Gene Ontology Consortium, 7 GO-biological processes that are related to netrophils (http://amigo.geneontology.org/amigo/search/ontology?q=neutrophils)

B) The C7 immunologic signatures from WHEI (rdata files)

My major questions are:

1) In the context of microarrays, especially for the first part of the specific GOs: fry would be more appropriate, or mroast ? Alternatively,

would mroast be more suitable for the second part with the many immunologic gene sets ?

2) My second issue, is more specific with the microarray platform and annotation:

in detail, the microarray platform is the Agilent SurePrint G3 Human GE v2 8x60k Microarray (Array Design A-MEXP-2320),

for which as no R annotation package was available, i have downloaded the latest gene symbol annotation from https://earray.chem.agilent.com/earray/

Thus, as both of the above approaches need Entrez Gene ids, how could i proceed ? as my expression matrix, has unique gene symbols in the rows ? Below, is a small code chunk from the final limma part:

class(final)
"EList"
attr(,"package")
"limma"

 dim(final$E)
23339   119

head(final$E)
     US84600244_253949426815_S01_GE1_107_Sep09_1_4
IRX1                                      4.979257
SAA1                                      7.548621
H19                                      13.150892
MBP                                       8.240486
SAA2                                      6.692976
CHGA                                      7.527782.....

condition <- factor(final$targets$UBE2D3.group,
levels = c("LOW.UBE2D3","HIGH.UBE2D3"))

design <- model.matrix(~condition)

fit <- lmFit(final,design)...

Thank you in advance,

Efstathios

1) With 7 particular GO terms, I would use mroast. Why not? roast is designed for focused gene set tests. fry is an approximation to mroast but, with only 7 terms, you may as well use roast itself.

For B) I would use camera.

2) Personally, I use alias2SymbolUsingNCBI() to convert gene symbols to Entrez Gene Ids and anything else I need. For example:

> Symbols <- c("IRX1","SAA1","H19","MBP","SAA2","CHGA")
> alias2SymbolUsingNCBI(Symbols, "Homo_sapiens.gene_info")
      GeneID Symbol                                    description
14710  79192   IRX1                            iroquois homeobox 1
5055    6288   SAA1                               serum amyloid A1
20753 283120    H19 H19, imprinted maternally expressed transcript
3388    4155    MBP                           myelin basic protein
5056    6289   SAA2                               serum amyloid A2
925     1113   CHGA                                 chromogranin A