So, I am trying to find DE genes in a cancer-normal dataset. I am using both DESeq2 and limma/voom method. Interestingly, I found many genes that are significance (adj p value <0.05) but the log2FoldChange sign is on the opposite. So, for example, in DESeq2 it is reported as down regulated but in Limma/voom result it is reported as up regulated. I check this because I am trying to inspect a gene that are well known in cancer but this inconsistent result confuse me. Below is the picture. The x axis is log fold change reported by limma/voom and y axis is the log fold change reported by DESeq2. Each data point represent a gene and all genes that are displayed here have padj-value <0.05. What do you think that cause this and which should I use? Majority of the results are quite consistent though.
Below is the plot for readcount per gene from salmon accummulated using tximport. I use this readcount data for both DESeq2 and limma/voom. The x-axis is the sample. Sample 1-7 are normal and 8-14 are cancer. So, it seems from readcount, the gene should be downregulated so DESeq2 is reporting the correct result.