Padjusted value = 1 in wilcox test

Question

[swathy]

Hi,

I am working on a treatment's bulk RNA-seq of responders and non-responders(5 responders,4 non-responders). P adjusted values are 1 when tried with the Wilcox test in Seurat, presto packages. But in Deseq2 and EdgeR(T-Test), there are significant FDR values. I have tried with FPKM values after normalization.

Thanks in advance.

Here is the code I used:

pvalues <- sapply(1:nrow(count_norm),function(i){                                                                                                
             data<-cbind.data.frame(gene=as.numeric(t(count_norm[i,])),conditions)                                                        
             p=wilcox.test(gene ~ conditions, data)$p.value                                                                                    
             return(p)                                                                                                                                                          
})

fdr=p.adjust(pvalues,method = "BH")

Answer 1

[ATpoint]

The Wilcox test is massively underpowered at low sample size. This, together with other reasons, is why specialized methods exist.

Don't use Seurat here. Use DESeq2 with the raw counts as input. That is statistically sound and gets reliable results. See its vignette at Bioconductor to see the recommended workflow.

[swathy]

Thank you. I will go with DESeq2 results.