I did my analysis using EdgeR and with DESeq2 with and without lfcShrink. At FDR<10% we have the following:
EdgeR - 17 genes
DESeq2 without LFC shrink - 37
DESeq2 with LFC shrink - 54
Only 10 genes overlap between all the 3 methods at FDR<10%. The fold change for EdgeR and DESeq2 without LFC shrink are almost the same for these 10 genes are high. But when the analysis is repeated with DESeq2withLFCShrink the fold changes for these 10 genes drop to < 2.
Our results show that EdgeR and DESeq2 without lfcShrink show very high fold changes (>10) which is highly unlikely in our study. On the other hand DESeq2 with lfcShrink is giving reasonable fold changes but the number of differentially expressed genes at FDR<5% or 10% is high compared to EdgeR.
Is there an explanation for why the fold changes are inflated for EdgeR and DESeq2 without lfcShrink?
Is there a lfcShrink in EdgeR?