Different gene "nomenclature" in the list of DEGS

Question

[Leite]

Hello everyone,

I did a microarray analysis and in my list of DEGS showed a mixed of gene nomenclature:

A_32_P16931
A_32_P196669
A_32_P721983
AA585242
AA702627
ABCA11
ABCC9
EFCBP1
ELN
ENST00000258457
ENST00000262042
ENST00000295989
ENST00000315707

I've seen in some published work that this mix is common in some lists of DEGs, however I wondered if this can influence in the GO/pathway functional enrichment or an interaction network?

Bets regards

Leite

[GenoMax]

How did that happen? You did not start with a consistent set of ID's or is this two different kinds of data merged for analysis?

[Leite]

I dont know, but I think i dont set a consistent ID

I found other papers with this mix of gene names: paper 1, paper 2 and 3.

This is the code I used for Agilent-014850 Whole Human Genome Microarray 4x44K G4112F:

rm(list=ls(all=TRUE)) 

library(Biobase)
library(GEOquery)
library(limma)

#Read in tab-delimited targets file
targets <- readTargets("targets.txt")

#load data into an RGList object
x <- read.maimages( files=targets, path=".", source= "agilent", green.only= T, columns= list( G= "gProcessedSignal") )

#Pre-processing
ex <- x$E
rownames(ex) <- x$genes$ProbeName
ex <- log(ex,2)
y <- x
y$E <- ex
y <- avereps.EList(y, ID=y$genes$ProbeName) #avereps_H.EList
y$E <- 2 ^ y$E 
y <- backgroundCorrect(y, method="normexp", offset=16)
y <- normalizeBetweenArrays(y, method="quantile") #normalizeBetweenArrays normalizeWithinArrays quantile 
#average replicate spots
y.ave <- avereps(y, ID=y$genes$ProbeName)

#Build the design matrix for the linear modelling function
f <- factor(targets$Target, levels = unique(targets$Target))
design <- model.matrix(~0 + f)
colnames(design) <- levels(f)

#Apply the intensity values to lmFit
fit <- lmFit(y.ave, design)
write.table(fit, file="fit.txt", sep="\t", quote=FALSE)

#Create a contrast matrix
contrast.matrix <- makeContrasts("D7D-Control", levels=design)

#Apply this contrast matrix to the modeled data and compute statistics for the data
fit2 <- contrasts.fit(fit, contrast.matrix)
fit2 <- eBayes(fit2)

#Output the statistics for the dataset and write them to disk
output <- topTable(fit2, coef=1, genelist=y.ave$genes, number=Inf, adjust.method="BH", lfc=1.5)
write.table(output, file="D7D-Control.txt", sep="\t", quote=FALSE)

Best regards, Leite

[GenoMax]

Here is the platform on GEO. It looks like you code must not have consistently chosen the primary column to work with. If you look at the table there are multiple blank cells in columns other than the first 2.

[Leite]

Hey genomax, thank you so much,

The platfmor on GEO is GPL4133, so my code have a problem? What I can I do?

Best regards,

Leite

[GenoMax]

One I linked above is a 2 year newer version of the platform you linked and was updated this year. Which tool were you planning to use for GO enrichment? Use the ID's that the tool works best with.

[Leite]

I'm using Reactome for pathways and PANTHER for GO.

So, Do you think I can have problems?

[GenoMax]

Tools may not like mixed identifiers. Did you check the annotation table to see if the mix of ID's you ended up with is because of missing values in columns?

[Leite]

Both platforms accepted the mixed data, some genes were not found, but I believe it is not because of the mix ID.

I'm thinking about manually filtering for uniprot ID or NCBI,

What you think about?

[GenoMax]

Output the agilent ID's of the DE genes and then cross-ref the table to see why you are ending up with mixed ID's.

You can filter the ID's manually but then you would lose a fraction of the data. Will it be significant?

[Leite]

I believe it would not make a significant difference, it would only make the table more organized.

Thank you.