ChAMP ebGSEA error

Question

[macsuejacques]

Hello,

I have been running the ebGSEA method for my methylation samples and it works fine with all other phenotypes that I have but for one particular one I keep getting the following error:

[ Section 2: Running Global Test Start ]

Applying Binary Model on Global Test. It could be very slow... Error in eval(family$initialize) : y values must be 0 <= y <= 1

My phenotype values are between -1 and 1, I've tried the same with absolute values but also does not work.

I am using the Beta values so they are between 0-1

I cant figure it our since my other phenotypes were definitely higher than 1 and all worked fine.

[Felix]

Encountering the same issue:

Running: myebGSEA <- champ.ebGSEA(beta=myNorm, pheno=myLoad$pd$Sample_Group, arraytype="EPIC")

Error: ebGSEA function require no NA in beta and pheno parameter.

[ Section 1: Generate Annotation Start ]

Extracting annotation from IlluminaHumanMethylationEPICilm10b4.hg19. Removing Non-CG probes out of annotation. Flat all genes on each CpG. Removing all duplicated CpG-genes. Annotation Prepared.

[ Section 1: Generate Annotation Done ]

[ Section 2: Running Global Test Start ]

Applying Binary Model on Global Test. It could be very slow... Fehler in eval(family$initialize) : y values must be 0 <= y <= 1 Zusätzlich: Warnmeldung: In alias2SymbolTable(flat$symbol) : Multiple symbols ignored for one or more aliases

Answer 1

[Josh]

(Answering this question because it's a top result on google when I encountered the same problem)

This issue occurs when myLoad$pd$Sample_Group is non-binary. It appears that the ebGSEA function requires binary phenotypes. To get around this, you could do something like this:

a_vs_b <- myLoad$pd$Sample_Group == 'a' | myLoad$pd$Sample_Group == 'b'
pheno_a_vs_b <- myLoad$pd$Sample_Group[a_vs_b]

myebGSEA <- champ.ebGSEA(beta=myNorm[,a_vs_b], pheno=pheno_a_vs_b, arraytype="EPIC")

where a and b binarize your phenotype values.