Entering edit mode
6.1 years ago
Daniel-X
•
0
Is it possible that GSEA results (top 20 gene sets with highest NES, or gene sets with FDR<0.05) are almost the same, expression datasets from different patients?
gene sets database: C2 all curated gene sets and C5 GO gene sets.
Am I fooled by the sequencing company?!
Please elaborate on the experimental setup and the analysis which was performed. Also, avoid abbreviations which may be common in your subfield but not for all of us.
GSEA: gene set enrichment analysis
NES: Normalized enrichment score.
C2 and C5 are both defined by the molecular signature database(MSigDB), totally almost 10k gene sets.
What about the experimental setup?
Did you perform GSEA patient by patient, all with the same disease?
Or GSEA was performed on groups of patients sharing a disease, comparing to another group of patients with same disease but different treatment?
Or GSEA was performed on groups of patients sharing a disease, comparing to another group of patients with another disease?
When you have doubt verify the results with other tools available for gene enrichment analysis.
thank u all for ur advice,
I did perform GSEA patient by patient, all with the same type of tumor, adjacent normal tissues (ANT) are available for each patient, expression data based on comparison between tumor and ANT from the same patient, after standard process of RNA-Seq data.
I carried out DAVID (gene annotation functional analysis) with my data too, the input of DAVID is just about gene names.
It seems the clusters from DAVID are identical too, GO term: "synapse", "Cell junction" rank at top of down-regulated pathways, "ribosome", "Ribosomal protein", "cytosolic ribosome" rank at top of up-regulated pathways in all patients, this does not make any sense...
I don't know the details of how you performed the differential expression analysis and the GSEA analysis, but in general, these analyses are performed group-wise: how the expression and / or metabolic pathways are affected on all tumor tissues samples vs all normal tissues samples (in the case of your experiment, taking advantage the paired nature of the design).