Entering edit mode
Hallo community,
I recently analyzed age x genotype interactions in my bulk RNAseq data using LRT in DESeq2, however, most of the interactions are no longer present when assessed by two-way ANOVA. Question: Would you consider two-way ANOVA more reliable than LRT in assessing interactions or vice versa? Any other suggestions on how to assess interactions described above would be greatly appreciated. Looking forward to your feedback
Enzo
Hi Michael,
Based on your body of work on RNAseq data analysis in DESeq2 (which i have reviewed extensively), it seems to me that LRT would be best suited for assessing interactions in my dataset. However, my boss was skeptical about the validity of these interactions on the basis that after testing several of them using 2-way ANOVA, they disappear! The reasoning is that 2-way ANOVA which may be used to check for interactions between 2 independent variables should yield similar results of the interactions identified by LRT. Does this make sense to you?
Enzo
I just don’t know how you’ve run an ANOVA on count data. If this were a valid approach then methods like DESeq2, edgeR, voom would not exist.