I am performing differential gene expression analysis and I am not sure what is an ideal cutoff or threshold for log2(Fold change). To my understanding, we only require acceptable cutoff for p-value, however, is there any criteria to decide threshold for fold change?
There is neither a common acceptable cut-off for the p-value nor for the LFC.
If a particular LFC is interesting depends on the gene, its function and on the biological system/sample you are looking at. For instance, a two-fold induction of a cytokine upon an infection can be uninteresting/negligible, but a 1.2-fold induction of a gene being expressed only in a couple of cells that drives a biological process can be a highly relevant finding. Similar for the p-value, what just gives a kind of a signal-to-noise ratio. A noisy expression can be interesting biologically.
I Agree with Jochen Wilhelm .
Further, even a small change in key transcription factors expression may alter the entire system. So this becomes very context-specific.
Large changes in low-level expressed genes may be less biologically significant than smaller changes in highly-expressed genes, so depends on your question and gene(s) of interest.
Well, I also agree with Jochen Wilhelm.
You need to know the reason why you need to get DEGs. Usually, we use FDR>=0.05 & |log2FC| >=1, finding specific gene sets to do enrichment analysis or other analysis. Sometimes if the DEGs are too much, you can try to let the threshold more strict, maybe >=1.2, 1.5 or 2. Also if the DEGs are too few, you can change the |log2FC|, maybe >=0. So, actually it depends on what you wanna do next. They are just numbers, you should consider more about the biological meaning behind the datasets.
- Badges
- Science topic
- Similar topics
- Biological Science
- Molecular Cell Biology
- Culture Cells
- Cell Line