Hi everyone, I am doing DGE analysis using limma-voom. I found that my logFC values are relatively small, ranging from approximately -0.10 to 0.07 (see column 2 below):
I should note that I imported the data from GEO Series Matrix File(s) and I might accidentally logged the processed logFC data in the matrix file.
Is there a specific instruction in limma to prevent the “log of a log” issue?
Many thanks!
Hi @bcgoh,
I am somewhat confused with the question. Do you suspect that you accidentally used wrong type of data, logFC, instead of counts? If yes, can you check the input data? Usually, logFC come in both negative and positive, but counts cannot be negative. If you put negative values into limma-voom input, I expect the job will fail.
The adjusted P-values are fairly high, consistent with small difference in expression level.
Kind regards,
Igor
Hi @igor
I think my Log2FC values are incorrrect, they shouldn’t have such small values. If you refer to the volcano plot below, you may notice that the Log2FC values (x-axis) of UseGalaxy generated data (left) has narrower range than the values from GEO2R (right). They are from the same dataset.
I suspect that I accidentally processed the “pre-processed data”, and I tried to set my normalisation method to “None (Do not normalise)” instead of “TMM”. However, the Log2FC values still appear the same. I wonder what could go wrong.
The input file was from Series Matrix File(s), GEO. I’ve attached the imported data for your reference:
Hi @bcgoh
The screenshot of the table does not provide any description of the data. Limma (limma-voom) uses count tables. Are you sure your data contains read counts? Reads counts are usually integers.
Consider following any established protocol for RNA-Seq analysis, e.g., Hands-on: 2: RNA-seq counts to genes / 2: RNA-seq counts to genes / Transcriptomics
Check the input data and compare it with your table. Note the difference.
Maybe talk to owner of the data or people who created it.
Kind regards,
Igor
