Dear all,
I have several GO lists derived from differentially expressed genes. The GO terms are almost identical across the lists, however, the number of a given GO term and FDR value are different. I was wondering how I can appropriately compare the GO terms among the lists. For comparing the number, I did a statistical test to find which GO terms are significantly different among the lists, and got some GO terms, please kindly let me know if it's a right approach? however, there are some GO terms that their number aren't different among the list, but their FDR are different, for example, please consider the below GO term:
Number in input list Number in reference FDR
list1 GO:0006820 68 266 7.40E-05 list2 GO:0006820 65 266 0.048
Please kindly tell me if this difference is important and has a biological meaning or interpretation? If your response is positive, could you please help me out how I can compare the given GO terms of several lists in term of FDR?
Thank you in advance
"For comparing the number, I did a statistical test to find which GO terms are significantly different among the lists, and got some GO terms, please kindly let me know if it's a right approach? "
I do not think it is necessary to compare GO terms accross different lists. Instead summarize the status of terms from different lists. As I mentioned in your earlier question, present your logFDR along with number of genes in individual lists.
C: Gene ontology comparsion between up- and down-regulated genes
This also depends on how many lists you are dealing with. Some times more lists become difficult to present graphically. In that case use simple table like this with addition of number of genes column.
Thank you for your comment. Actually, I faced with about 500 GO terms that significantly enriched in the up-regulated genes in both types of a cell under comparison. So, it's a bit difficult, at least for me, to compare them, that's why I did a statistical test to find differences between these GO terms belong to two types of cell. Regarding my last issue about FDR, could you please share me your opinion? Again thanks
500 GO terms questions the reliability of your input gene lists.
1) Approximately how many genes were in the input list.
2) If it is less and trustworthy, are you sure that you chose proper background specific to your organism while performing enrichment analysis?
3) I guess you might have to reconsider filter criteria for selecting genes (if you get more genes).
4) If number of genes and the background selection is fine. In addition to FDR, introduce number of genes or percentage of genes detected from particular function as filter criteria while choosing the GO term to be significant.
There are about 2500-3000 significantly differentially expressed genes (FDR < 0.05 and logFC cutoff 1) derived from an RNA-seq analysis of a non-model plant. I used the assembled transcriptome that annotated against Arabidopsis proteome as background reference. Regarding your suggestion on item 4, my issue is not a large number of GO term as I saw the similar situation in papers. Also, just a few of these GO terms are present in one list and absent in another, which these GO are my interest. However, I wanted to compare the common GO terms between list, that may ignore it. Sorry, again comparison!. I found that several overrepresented GO terms are common between up- and down-regulated genes, I have some interpretation for overrepresented GO in the up-regulated gene list, but when this GO term is also found in the down-regulated gene, this interpretation is not useful. Please kindly tell me how I can explain this issue and handle such results? if can I consider the related FDR of a given GO term between up and down-regulated gene to decided if the Go term enrich in the up or down-regulated gene?
Thanks in advance for your help.