thank you!
Hi,
I'm performing Gene Ontology (GO) enrichment analysis on differentially expressed genes (DEGs) from ATAC-seq data, using R (clusterProfiler). After separating my DEGs (DEseq2) into up-regulated and down-regulated gene sets based on log2 fold change and adjusted p-value thresholds, I noticed that the GO pathway analysis for both sets returns similar or identical enriched pathways.
Has anyone encountered this issue? What could be causing the same GO terms to appear in both up- and down-regulated gene sets? As output from DEseq2, the genes from the two list cannot be shared.
1 answer
There is absolutely no reason why you cannot have an enrichment of both up and down regulated genes in the same pathway. Consider a pathway with 20 genes. 10 of those genes are up regulated and 10 downregulated. Consider also that you have 1000 significantly upregulated genes and 1000 significantly downregulated genes in total. 1000/20000 or 5% of all genes are DE up and 5% DE down. 50% of pathway genes are up, which is a 10 fold enrichment (50%/5%), and 50% of pathway genes are down, which is also a 10 fold enrichment (50%/5%).
We must be careful when interpreting the results of enrichment analyses. An enrichment of upregulated genes does not mean that activity from the pathway is upregulated. For example, it could be inhibitors of the pathway that are upregulated. In your analysis, its possible that pathway activators are upregulated and pathway inhibitors are down regulated. Or it could be the other way around. Most likely it is not as clean cut as that though - biology moves in mysterious ways, and most genes can't be cleanly divided into activators and repressors.
Remember that enrichment analysis is a purely hypothesis generating technique. It tells you what subjects/functions you should pay more attention to. It doesn't tell you what is happening to them.
Log in to answer this question.
ATAC-seq measures chromatin openness, not expression, so you shouldn't call these genes differential expressed genes. They are differentially accessible genes. Further, it is likely that any change in openness is not uniformally distributed across the genes. At least in my experience, genes that are upregulated at the expression level to not neccessarily increase in openness at the chromatin level and vice-versa.